7

# Research

Stacy A. Weil, RDH, MS

Objectives

Upon completion of this chapter, the student will be able to:

• Differentiate between the hypothesis and the null hypothesis of a research study.

• Explain the importance of the scientific method in research.

• Define a population and a sample as related to research.

• Discuss sampling techniques and their uses.

• Discuss the difference between the independent and dependent variables.

• Use the terms mean, median, and mode to express the results of data collection.

• Define the terms continuous data and discrete data and their respective scales of measurement.

• Discuss the uses of various statistical techniques.

• Use different types of displays to exhibit data.

• Explain the difference between type I and type II errors.

• Define probability and statistical significance.

• Express the importance of evaluating dental literature.

• Explain the criteria for reviewing scientific literature.

• Review a scientific journal article relating to dentistry.

Key Terms

Hypothesis

Null hypothesis

Calibrated

Validity

Reliability

Interrater reliability

Intrarater reliability

Population

Parameter

Target population

Sample

Statistic

Pilot study

Random sampling

Stratified sampling

Systematic sampling

Purposive (judgmental) sampling

Convenience sampling

Experimental group

Control group

Independent variable

Variable

Dependent variable

Data

Discrete data

Continuous data

Nominal scales

Ordinal scales

Interval scales

Ratio scales

Descriptive statistics

Inferential statistics

Mean

Median

Mode

Range

Variance

Standard deviation (SD)

Correlation

Parametric

Normal distribution

t-test

Analysis of variance (ANOVA)

Nonparametric

Chi-squared test

Power analysis

P values

Type I alpha (α) errors

Type II beta (β) errors

Refereed

Abstract

## Opening Statements

### Questions in Research

• How does a public health team decide that fluoridation in a community’s water supply will reduce the incidence of new carious lesions?

• Exactly how much fluoride is required to add to the water to achieve a therapeutic effect without a toxic reaction?

• How would officials determine that dental hygienists might improve the quality of life in older people if they were able to work independently in extended-care facilities?

• How do communities decide to spend money on a program providing dental care to individuals with human immunodeficiency virus (HIV) infection instead of to children with special disabilities?

• Where can dental hygienists readily find employment, and what salary can they expect to earn?

• What is the effectiveness of school based sealants in managing caries?

## Questions and Answers in Research

Although dental hygienists may seek the answers to the questions in the Opening Statements, patients may have other concerns such as the following:

• Does a particular mouthrinse really reduce plaque buildup?

• Which brand of toothpaste is best?

• Can nonsurgical periodontal therapy provide results comparable to those of a surgical procedure?

Even though students may commonly learn the answers to these questions from instructors or colleagues, it is important to understand where to find reliable answers to these questions independently and to understand the process that provides these answers.

Research via the scientific method is the basis from which these answers are produced. Manuscripts published in reputable scientific journals disseminate the results of independent research. To determine whether the information contained therein is indeed reliable and valid, certain knowledge and skills must be a part of the repertoire of every competent dental hygienist practicing in the dental community. This chapter provides a basic outline of what research entails and a method of evaluating the results of that research.

## The Scientific Method and Development of a Research Problem

Understanding the basics of research entails gaining an appreciation for the components of a good research study, that is, understanding how a research idea is formulated, how a study is designed and executed, and how the resulting data are critically evaluated so that one can infer appropriate conclusions. Research can be thought of as a search for truth and the knowledge gained from this search. A true definition of research is a systematic inquiry that uses orderly scientific methods to answer questions or solve problems.1

The discoveries provided by research may lead to new knowledge or to the revision of existing knowledge. Dental research involves a systematic search for knowledge about issues relevant to the profession. To increase the chance that research will be valid, reliable, and relevant, the scientific method—a series of logical steps starting with the formulation of a problem—is employed (Box 7-1).

BOX 7-1   The Scientific Method

• Formulation of a problem (asking the question)

• Formulation of a hypothesis (a proposed answer to the question)

• Collecting the data (finding existing information related to the question, as well as gathering of your own information)

• Analysis and interpretation of the results

• Presentation of the results

• Formulation of conclusion (relationship of results to hypothesis)

### Formulation of a Problem (Asking the Question)

The first step in beginning a research study is the formulation of an idea. The idea is usually formed from a question that has been raised by a researcher. The question may arise from a very simple observation or thought. For example, during the clinical phase of dental hygiene education, participants might debate about the following:

1. Which areas of the mouth are most difficult to probe accurately?

2. What are the effects of diet on periodontal disease?

3. How can adequate oral health care be maintained by physically and mentally challenged people?

From simple notions such as these, a research question or problem can be formulated.

Examples of a research problem formulated from the previous questions might be as follows:

1. Which quadrant in the human dentition is least accurately probed by the second-year dental hygiene student at University X when using the Periodontal Screening Record (PSR) method of probing?

2. What is the percentage of calories from carbohydrates in the diet of patients exhibiting class II periodontal disease?

3. What effect does modifying the brushing techniques of disabled patients in long-term care facilities have on the gingival bleeding index of these patients?

The research problem should be kept as simple and concise as possible. A successful study often depends on an uncomplicated research design, which results from simple questions.

### Formulation of a Hypothesis (a Proposed Answer to the Question)

After a research question is formulated, the next step is the development of a hypothesis, a statement that reflects the research question. The hypothesis is stated in positive terms that represent the researcher’s prediction or opinion. An example of a hypothesis for the question “Which quadrant in the human dentition is least accurately probed by the second-year dental hygiene student at X University when using the PSR method of probing?” would be as follows: Second-year dental hygiene students at University X using the PSR method are most inaccurate when probing the distal lingual surface of teeth in the upper right quadrant of the mouth.

The research statement is often expressed as a null hypothesis, which assumes that there is no statistically significant difference between the groups being studied. An example of a null hypothesis for the preceding question would be as follows: Second-year dental hygiene students at University X show no difference in the accuracy of probing any tooth in the mouth when using the PSR method.

Once the hypothesis has been formed, data can be collected to prove or disprove the statement.

### Collecting Data (Finding Existing Information and Gathering Information Independently)

After a research idea and a hypothesis are initially identified, the relevant available literature is reviewed. By examining an area of general interest, the researcher may be inspired to create a research question to resolve unknown or unexplained portions of the area of interest. Alternatively, when a general topic has been selected, a literature review may help to bring the topic into sharper focus. Emulating the accepted research designs that have been previously validated by others, the researcher can then design a study to evaluate the idea. Analysis of the literature is described later in this chapter.

After review and analysis of the available literature, the researcher can plan how the study will be conducted and how the data will be collected. Many different techniques can be used to collect data (see Chapter 3). During data collection, it is important to use calibrated instruments that are both valid and reliable. When examiners are involved in data collection, it is imperative that they be calibrated (i.e., in agreement with a set standard of performance for the data collection). For example, if one examiner notes caries on the occlusal surface of a first molar, a second examiner should also be able to note caries on the occlusal surface of the same first molar.

Validity is concerned with gathering data that have been intended to be collected. For example, if two calibrated dentists are examining children for occlusal caries on first molars, both dentists must examine and record caries only on the occlusal surfaces and only on first molars. They are examining the correct surfaces and collecting data that were intended to be collected.

Reliability refers to the consistency and stability of the data. The data are reliable if the examiners are calibrated and can reproduce the results. Both examiners must find the same three occlusal caries on first molars in child No. 1. If they examine child No. 1 an hour later, they should still find the same three occlusal caries as detected previously.

A rater is the person who is collecting the data or making an assessment during a research experiment. The reliability of the data is important, and raters must be consistent both with their own findings, as well as each other’s assessment of the same situation. The terms interrater reliability and intrarater reliability are used to describe this consistency between and within each rater. Interrater reliability refers to the extent that two or more raters obtain the same result when using the same instrument to measure a concept. Intrarater reliability or consistency refers to the same rater making the same assessment on two or more occasions.

In developing the plan for conducting the study, it is important to identify the characteristics of the group involved in the study. Group characteristics are defined by such terms as population, sample, experimental, and control. The term data, or information, collected from the study group is defined and used in different ways. These terms and their relationship to the collection of research data are explained next.

#### Population and Sampling

##### Population

Population can be defined as the entire group or whole unit of individuals having similar characteristics from which the results of an investigation can be inferred.2 In regard to numeric characteristics of the population, the term parameter is used.

Populations can be very large or very small, depending on the topic to be studied. For example, in the first research question (which quadrant in the human dentition is least accurately probed by second-year dental hygiene students at University X using the PSR method of probing?), one can infer that the total population consists of second-year dental hygiene students. It would be optimal to examine each of these students to arrive at the answer. The second-year dental hygiene students are also known as the target population, or the population from whom the information is being collected. Because of time constraints, lack of resources, or financial issues, however, it may be decided that a smaller group within this group can provide the researchers with a significant result.

##### Sampling

Taking a portion of the population is known as sampling. A sample is a portion or subset of the entire population that, if properly selected, can provide meaningful information about the entire population. When one is discussing numeric characteristics of samples, the term statistic is used.

Samples can be large or small and are chosen to most appropriately reflect the research being done. A large sample usually provides the most accurate representation of the population and increases the exactness and accuracy of the data collected. Occasionally a small sample may be used, as in the case of data collection for a pilot study, or trial run, done in preparation for a major study.

The importance of using a sample becomes obvious in regard to our research question. Suppose there is an urgent need to refine teaching techniques for probing and there is not enough time to adequately assess each second-year dental hygiene student’s probing at University X. The researcher may thus decide to use only a portion, or sample, of the first-year class.

If it is decided that a sample of the population is to be used, different techniques are used to choose the sample. There are several types of sampling.

###### Random Sampling

The method of sampling that provides the most external validity, or degree to which the results of the study can be generalized to settings other than the one included, is called random sampling. This method provides a sample in which each member of a population has an equal chance of being included and is the procedure of choice because it prevents the possibility of selection bias by the researcher.

As an example, assume that all second-year dental hygiene students at University X have been given a number. There are 50 students in the second-year class, and it is determined that a sample of 10 students will participate in the research. Numbers 1 to 50 would be written on separate slips of paper and placed in a container. From the container a number is drawn. The number is noted as one of the 10 students to be used in this sample. That number is then placed back in the container, and another selection is made until the list of 10 numbers is complete. The importance of placing each number selected back into the container is to preserve a true random selection from 50 numbers. Numbers that are drawn more than once are put on the list only once and are again placed back in the container.

###### Stratified Sampling

What if probing discrepancies are unique to the university’s dental hygiene program? A random sample may not accurately assess the problem for all second-year dental hygiene students; it may be necessary to include students from other universities as well in a method called stratified sampling. Subdivisions of a population with similar characteristics, such as second-year dental hygiene students attending different dental hygiene programs, are called strata. The random selection of subjects from two or more strata of the population is another way of defining stratified sampling.

###### Systematic Sampling

Another form of sampling, systematic sampling, involves the selection of subjects by including every nth person in a list. For example, if a researcher had a list of second-year dental hygiene students by number and chose every odd-numbered person, he or she would be sampling the population systematically. In this case, unless the list is in random order, not every person may have an equal or random chance of being selected; thus systematic sampling may not be considered a true random sample.

###### Purposive (Judgmental) Sampling

If an instructor who most often works with students who are learning probing techniques chooses the sample, it is easy to imagine that a great deal of bias may be introduced into the study. Purposive or judgmental sampling provides a sample, through personal judgment, of subjects who would be most representative of the population.

###### Convenience Sampling

A convenience sample may also introduce bias. Selection of a sample through convenience sampling provides a group of individuals who are most readily available to be subjects in the study. For example, a researcher conducting the probing study might enroll only individuals from the researcher’s university as subjects.

#### Experimental and Control Groups and Variables

##### Experimental and Control Groups

After the sample population is selected, subjects may be divided randomly into experimental and control groups, which are used to answer a research question posed when an experimental treatment or manipulation is imposed on the research setting. The experimental group is the sample group in a study that receives the experimental treatment or intervention. The control group is the group in a study that does not receive the experimental treatment or intervention. The control group provides the baseline against which the effects of the intervention on the experimental group can be measured.

For example, assume that the study on probing accuracy found an area of the mouth in which inaccuracy predominated. At this point, one might choose to conduct another research study, introducing a new method of probing instruction that would focus on the area of the mouth in which most inaccuracies occur. This new method of teaching would be administered to the experimental group, whereas the control group would continue to receive the traditional method of instruction.

##### Variables

The experimental treatment or intervention that is imposed on the experimental group can also be called the independent variable. A variable is a characteristic or concept that varies, or is different, within the population under study. The independent variable is controlled or manipulated by the researcher and is believed to cause or influence the dependent variable. The dependent variable is thought to depend on or to be caused by the independent variable. It is the outcome variable of interest.

In the case of the research question, which involves teaching a new probing technique to the experimental group, the dependent variable would be probing accuracy. The independent variable would be the probing technique that is being taught. Other variables not related to the purpose of the study are uncontrolled variables and may influence the relationship between the independent and dependent variable. To increase internal validity, it is important to control for extraneous variables. Internal validity refers to the fact that it is the experimental treatment or independent variable that is responsible for the observed effects and that these effects are not due to extraneous variables. A good research study controls for extraneous variables through research design or through statistical procedures.3 By understanding the previous terms and definitions, one can implement a method of data collection, or the gathering of information, to address a research problem.

#### Data

Pieces of information, such as numbers collected from measurements and counts obtained during the course of a research study, are known as data. Although the concept of data itself may seem fairly straightforward, there are different types of data and different ways to measure data (Table 7-1). Two types of data are as follows:

Table 7-1

Types of Data

 Types of Data Characteristics Examples Related Scales of Measurement Appropriate Data Display Discrete Limited set of values, represented as whole numbers, qualitative Hair color, number of times a person brushes, DMF teeth Nominal (named categories; e.g., male or female) Ordinal (same as nominal but categories in order; e.g., stages of cancer) Bar graph Continuous Particular value within a range, variables along a continuum, quantitative Temperature, test scores, time Interval (same as ordinal plus equal distance between variables but no zero; e.g., Fahrenheit) Ratio (same as interval plus equal distance between variable with zero; e.g., height, weight) Histogram

DMF, Decayed, missing, filled.

1. Discrete data have only one of a limited set of values and are counted only in whole numbers. Discrete variables may include things like hair color, gender, political preference, and number totals, such as how many times a person brushes his or her teeth or the number of decayed, missing, or filled teeth. These data are considered to be qualitative in nature.

2. Continuous data are measurements made from a particular value within a defined range. Variables along a continuum, such as temperature, scores on a test, and time, are continuous data. These data are considered to be quantitative in nature.

Different scales of measurement are used for discrete and continuous data. Discrete data can use nominal or ordinal scales of measurement. Continuous data use interval and ratio scales of measurements.

Nominal scales of measurement consist of named categories with no order. For example, females may be placed in category A and males in category B.

Ordinal scales of measurement consist of categories of variables in which the categories are in order, but there is no equal or defined distance between them. For example, cancer staging for tumors is grouped into four stages designated by Roman numerals I to IV. In general, stage I cancers are small localized cancers that are usually curable, whereas stage IV usually represents inoperable or metastatic cancer. Stage II and III cancers are usually locally advanced and/or with involvement of local lymph nodes. It is known that type II is worse than type I and that type III is worse than type II, but each type of cancer is slightly different, making it difficult to define precisely for all cancers.4

Interval scales of measurement have equal distance between variables, but there is no true zero point (i.e., temperature on a Fahrenheit thermometer).

Ratio scales of measurement have equal intervals between the variables, but there is a meaningful zero point (i.e., height and weight).

As listed in the following, each scale of measurement takes on the characteristics of the previous one, making ratio the most powerful measurement:

Nominal: Named categories only

Ordinal: Same as nominal plus categories are in order

Interval: Same as ordinal plus equal intervals between categories

Ratio: Same as interval plus meaningful or true zero point

After the data have been collected from a chosen population and variables have been defined and measured, it is time to proceed to the next step of the scientific method, presenting and interpreting the data and presenting the results.

### Data Analysis and Presentation of Results

#### Statistics

Statistics is a science that provides a way of processing numbers or analyzing the data that have been collected. Statistics may be used to describe, analyze, and interpret the numbers collected in the data. The purpose of statistical analysis is to make an inference or an assumption about a population.5 Two types of relevant statistics are as follows:

1. Descriptive statistics are used to describe and summarize data. Their objective is to communicate results, without generalizing beyond the sample, to any population. Some ways in which results are communicated are through (1) measures of central tendency (mean, median, and mode) and (2) measures of dispersion.

2. Inferential statistics (see later section) are used to apply information from the sample to a larger population.

#### Measures of Central Tendency

Measures of central tendency are used to describe the sample based on the data gathered. They include components such as the mean, median, and mode.

The mean is the average of the group. It is a sum of all the values divided by the number (n) of items and is statistically noted as . The mean is calculated in the following manner:

mean=∑x¯n

The positive aspect of the mean is that it includes the value of each score; the negative aspect is that it can be affected by any extreme scores and may not give a true average. For example, a test is administered; 10 people in the class score an 85 and 2 people score a 30. The class average becomes approximately 76, which is not a true representation of the class scores.

The median represents the exact middle score or value in an ordered distribution of scores; it is the point above and below which 50% of the scores lie. When the total number of scores is even, the sum of the two middle scores, divided by 2, provides the median.

Unlike the mean, extreme scores do not affect the median. In the previous example, the median score would be 85. However, it is not difficult to imagine what would happen if scores were not evenly distributed and the median were used as an example of the middle score. The information provided in this case may not demonstrate a true midpoint for the class.

The mode is the score or value that occurs most frequently in a distribution of scores. Once again, the mode for the preceding example would be 85. The distribution of scores may be unimodal, bimodal, or multimodal, or there may even be no mode.

Figure 7-1 presents the mean, median, and mode of a group of test scores. Figure 7-2 illustrates a symmetric distribution in which the mean, median, and mode are the same; it also shows skewed curves, where the mean and median are located to the left and right of the mode.

#### Measures of Dispersion

In addition to the measures of central tendency (mean, median, and mode), measures of dispersion (also known as measures of variation) may also be used to describe data. Measures of central tendency provide a first step in the measure of distributions. Occasionally, however, a measurement is also desired to determine how far scores differ from the mean.

The most obvious way of measuring the dispersion of data is the range, which measures the difference between the highest and lowest values in a distribution of scores. If the median is defined as the point where 50% of the scores are lower and 50% of the scores are higher, the location of the 0 and 100th percentile would define the range. More commonly, the value of the range is used to define the 5th and 95th percentiles. These are the points that 5% of the subjects may be below or 95% of the people may be below. If a person has taken a test and is in the 95th percentile, 95% of the people have scored lower than that person has.

Another measure of dispersion, called variance, is a method of ascertaining the way individual variables are located around the mean. Variance is most often used to measure interval and ratio variables.

A first step in the calculation of variance is to determine the average deviation, which can be derived through calculating the difference between each point of data and the mean, adding the answers together, and dividing by the total amount of data points. The main problem with calculating the average difference is that there are as many negative data points as positive; this situation ultimately results in an answer of zero.

A way around this problem is to square each data point so that all terms are positive. Therefore the difference between each data point and the mean squared, summed, and divided by the total amount of data points results in the average squared deviation, also known as variance.

Variance is simply a step in determining the standard deviation (SD) and is equal to the square of the SD. The formula for the SD is as follows:

SD=sum of (data point−mean)2No of values

It makes sense, then, that the farther away the data points are from the mean, the greater the variance and SD. Box 7-2 presents the calculations to find the range, variance, and SD of student test scores.

BOX 7-2   Range, Variance, and Standard Deviation of Student Test Scores

Student Test Scores

 Number of Students Score 6 60 3 30 3 100 4 45 4 95 5 90 5 50

Range is the difference between highest and lowest score: 100 − 30 = 70.

Variance is the average deviation or spread of scores around the mean.

The variance is calculated as (individual score − mean)2/# of scores.

(60 − 67)2 = 49

(30 − 67)2 = 1369

(100 − 67)2 = 1089

(45 − 67)2 = 484

(95 − 67)2 = 784

(90 − 67)2 = 529

(50 − 67)2 = 289

49 × 6 (# of scores of 60) = 294

1369 × 3 (# of scores of 30) = 4107

1089 × 3 (# of scores of 100) = 3267

484 × 4 (# of scores of 45) = 1936

784 × 4 (# of scores of 95) = 3136

529 × 5 (# of scores of 90) = 2645

289 × 5 (# of scores of 50) = 1445

All above summed = 16,830

16830 ÷ 30 (total number of scores) = 561

561 is the variance

Standard deviation is the positive square root of the variance:

The square root of 561 = 23.7 (standard deviation)

Box 7-3 demonstrates the calculations from data collected in a community research project involving unwed teenaged mothers and their knowledge of baby bottle tooth decay. Twelve mothers are in this study group. The mothers’ scores from the pretest range from a low score of 25% to a high score of 70%.

BOX 7-3   Calculations of Test Scores on Baby Bottle Tooth Decay

 Subject (mother) Score 1 45 2 45 3 45 4 30 5 35 Median = 45% 6 25 Mode = 45% 7 40 Mean = 48% 8 50 9 60 Range = 70 – 25 10 65 = 45 11 70 12 70 Variance = 210

#### Correlation

After we determine the previous information about our data, we may want to next determine whether any correlation exists between the variables. Correlation is a statistical method for determining whether a variation in one variable may be related to a variation in another variable. For example, height and weight often show a correlation because taller people usually have a higher weight than shorter people. Age and periodontal disease may have a correlation because older people may have higher incidence of periodontal disease than younger people.

The technique used to determine correlation depends on the type of variable being explored. Different techniques are used for discrete or continuous variables. The measurement scale may also influence the technique. For example, nominal, ordinal, interval, and ratio scales of measurement are all calculated slightly differently.

The results of the calculation for correlation show either a negative or positive relationship. When the relationship is positive, it is predicted that as the value of one variable increases, the other also increases. Perfect positive correlation is shown by +1.0. An example would be the finding that, in the research with unwed mothers, the more time the examiner spends with them on patient education, the greater the increase in the mothers’ knowledge. Table 7-2 presents a correlation of increased education with increased knowledge. Figure 7-3 demonstrates a graphic display of the same positive correlation of the two variables.

Table 7-2

Correlation of Test Scores and Hours of Education

 Group Number Group Type and Pretest Scores Hours of Education Average Posttest Score Increase (Points) % 1 Four mothers with an average pretest score of 50 2 60 2 Four mothers with an average pretest score of 50 4 70 3 Four mothers with an average pretest score of 50 6 80 4 Four mothers with an average pretest score of 50 8 90

In contrast, a negative correlation shows an inverse relationship between variables. Perfect negative correlation is shown by −1.0. Figure 7-4 shows that a diet including six servings of fruits and vegetables each day decreases the incidence of certain cancers.

In review, a perfect positive correlation is noted as +1.0 and a perfect negative correlation showing an inverse relationship is noted as −1.0 Although a perfect positive or perfect negative correlation may occur, it may be possible to find no relationship at all; this instance is noted as 0.0. Some literature states that generally a correlation coefficient above .70 is considered satisfactory.1 In each study, the nature of the variables and the numbers involved in the comparison must also be considered along with the correlation coefficient in determining what is a significant relationship. The closer the relationship is to +1.0 or −1.0, the more perfect, or stronger, the correlation.

#### Presentation of the Data

In addition to the data displayed in the previous graphs, other types of data display include the following:

A bar graph is most often used to display nominal or ordinal data that are discrete in nature. With the use of data that may be gathered from the study on teenaged mothers and baby bottle tooth decay, one can create a bar graph to present the data pictorially (Figure 7-5).

A frequency polygon is used to represent data that are continuous in nature. An example from the study with teenaged mothers is depicted in Figure 7-6. The figure contains dots connected to straight lines to present the frequency distribution of the data. In this case the frequency polygon demonstrates how many times per week the mothers brush their children’s teeth.

A histogram, although a type of bar graph, is used most often to represent interval or ratio scaled variables that are continuous in nature (Figure 7-7). The bars in a histogram are of equal width and touch each other to indicate that the data are being presented on a continuum.

#### Inferential Statistics

In addition to including descriptive techniques to describe data, inferential statistics may also be used. Whereas descriptive statistics are used to determine information only about the sample being studied, inferential statistics seek to determine a generalization between the sample studied and the actual population. The larger the sample size, therefore, the greater the number of generalizations that can be made regarding the population. Depending on the type of data collected, the use of inferential statistics may include either parametric or nonparametric statistical techniques. Inferential statistics are based on the assumption that sampling is conducted randomly, although that is not always the case.

#### Parametric Inferential Statistics

Parametric statistics are used when the data include interval or ratio scales of measurement. Parametric techniques work best when the sample is large and randomized and the population from which the sample is taken is normally distributed. In a normal distribution, 50% of the values lie on the left half of the distribution, and 50% lie on the right half. A normal distribution assumes that approximately 68% of the population fall within one SD of the mean, approximately 95% fall within two SDs of the mean, and 99% lie within three SDs from the mean.6 The plotting of these data on a graph results in a bell-shaped curve (Figure 7-8).

##### t-Test

Parametric statistics are calculated by means of several different methods or tests. One of the most common is the t-test, or Student’s t-test, so named for the man responsible for development of this technique. The t-test is used to analyze the difference between two means. It provides the researcher with the difference between treatment and control groups or groups receiving treatment A versus treatment B.

When the test is used on a single group that yields pretreatment and post-treatment scores, it is known as a t-test for dependent samples. This test analyzes data when only one independent variable is tested. For example, a researcher may want to examine the difference in blood glucose levels of diabetic subjects before and after treatment with a new diet. Assuming that all of the subjects were the same age and had the same degree of disease present, the t-test for dependent samples can be used to determine pretreatment and post-treatment scores.

The independent t-test determines differences in the mean between two independent groups such as an experimental and control group or males versus females. This test is used when only one or two independent variables are tested. Most often, data involving interval or ratio scales of measurement are statistically analyzed with an independent t-test. For example, a study might investigate the effect of a new toothbrushing method on gingivitis. The subjects would be randomized into two groups: (1) a control group receiving no instruction and asked to use their normal brushing method and (2) an experimental group asked to practice a new method of toothbrushing. The instrument may be a gingival bleeding index that is conducted before the start of the study and again 8 weeks later. The hypothesis is that the new method of toothbrushing will decrease the gingival bleeding index of patients with gingivitis.

##### Analysis of Variance

Another commonly used test for parametrics is analysis of variance, or ANOVA. This test allows comparison among more than two sample means and compares interactions among the variability in the multiple sample groups with the variability within the groups.

A simple example is the comparison of the many brands of desensitizing toothpaste. Five brands of toothpaste that claim relief of tooth pain caused by sensitivity have been located. A group of subjects is assembled, and each subject is given a different toothpaste disguised in a plain white tube. Each subject is asked to use this tube and is given other tubes to use until all of the various toothpastes are used. Subjects are asked to rank pain relief from tooth sensitivity on a numeric scale of 1 to 10 for each toothpaste used. The scale may look like the one in Table 7-3.

Table 7-3

Data Used in Analysis of Variance (ANOVA) Testing

 Patient Number TP 1 TP 2 TP 3 TP 4 TP 5 1 6.0 5.0 4.0 5.0 2.0 2 5.0 5.0 3.0 4.0 3.0 3 7.0 5.0 4.0 5.0 3.0 4 5.0 4.0 4.0 3.0 4.0 5 6.0 5.0 5.0 4.0 4.0

TP, Toothpaste brands 1 to 5.

Key: 1.0-10 = pain relief (10 is maximum relief).

ANOVA allows the dental hygienist to compare each toothpaste used and the pain relief experienced by the patients. It also allows a comparison of the different responses from each patient for each individual toothpaste brand. ANOVA may yield information about which of the brands is actually more effective and how each brand compares with another. In essence, ANOVA compares variability within groups with variability between groups.

The t-test, or Student’s t-test, and ANOVA are just two of many tests that may be used in statistical analysis of data. These tests are perhaps the most common parametric inferential statistical techniques.

No matter which technique is chosen, the sample should be randomly and independently selected from normal populations. Although these tests provide more detailed information of the interaction of variables, the research should include means and SDs in the report for accurate assessment of the tests’ magnitude.

#### Nonparametric Inferential Statistics

Besides parametric tests, nonparametric inferential statistical techniques may be selected. Nonparametric techniques are most useful for data to be measured on the nominal or ordinal scale. Remember, nominal or ordinal data are “qualitative,” and although numbers may be included, these numbers are derived more subjectively than numbers associated with quantitative data. Nonparametric tests involve fewer assumptions about the population. The sample size may be small, and variables are discrete.

The most commonly used nonparametric test is the chi-squared test. This test may be used to analyze questionnaire data and to determine whether a relationship exists between two variables. The chi-squared test is used to examine the differences between observed and expected frequencies.

#### Determining Statistical Significance

Statistical tests provide researchers with an idea of what the data they have collected say about the sample and perhaps what the data imply about the population from which the sample was drawn. An important factor in research is determining the statistical significance of the data. Statistical significance is a way of indicating that the results found in an analysis of data are unlikely to have been caused by chance; more likely, the results have been caused by the independent variable. Using too small or too large a sample may influence the statistical significance. Typically, the use of too small a sample (less than 30) provides too little information to make generalizations about the populations and to create any significance of results.

##### Power Analysis

Determining how many subjects are needed to provide significance is called a power analysis, which is calculated according to a specific statistical formula based on what the researcher hopes to observe in most of the subjects. The power of a study, or its ability to detect relationships among variables, is directly related to sample size, the definition of the independent variable, and the precision with which the study is planned and conducted.7 When too large a sample is used, the effects may be statistically significant but clinically of no consequence.

The true importance of determining statistical significance is that the greater the significance, the more statistical inference can be made regarding the population from whom the sample was taken. That is, statistical significance reinforces our ability to generalize the conclusions we make about our study population to a larger population, perhaps even to the “general population.”

A major issue in regard to statistical inferences is that every measurement taken from the sample being researched has some degree of error. Researchers may describe the possibility of error or lack of error in various ways.

##### Confidence Intervals

The term confidence interval refers to how researchers describe the probability of the statistical results being correct. A confidence interval indicates a range of values within which the parameters of the population have a probability of lying. Researchers usually use a 95% to 99% confidence interval. A 95% confidence interval indicates a probability that the researcher is wrong 5 times in 100. By using a 99% confidence interval, the researcher may be wrong 1% of the time; however, increasing the confidence interval also decreases the specificity of the data. Therefore, when the confidence interval is 95%, there is a 5% chance that the observed results or differences between study and control groups are due purely to chance and not a true difference caused by the independent variable.

##### P Values

Researchers also use P values to describe statistical significance. The P value states how likely it is that the study could have come to a false scientific conclusion. P values are calculated according to the sample size, the difference between the means of the control and experimental group, and the SD of the distribution. The smaller the P value, the more significant the findings of the study are considered.

A normally acceptable P value is P < .05. Results with a P value at less than .05 are generally considered statistically significant and provide the basis for rejection of the null hypothesis. P < .05 means that the results were due to chance only 5 times in 100. P values of approximately .01, .001, and lower increase the significance of the study.

#### Descriptive or Inferential?

• Last semester, the heights of students at the college ranged from 5 feet to 6 feet.

This is an example of a descriptive statistic because the data from the sample is exact (height of students) and there is no generalization applied to the group.

• Flossing can help prevent periodontal disease.

This is an example of an inferential statistic because the statement is a generalization of a larger population based on an interpretation of research data from a smaller study.

Table 7-4 compares descriptive statistics with inferential statistics.

Table 7-4

Types of Statistics

 Types of Statistics Characteristics Consider Measurement/Statistical Techniques Descriptive Describe and summarize data in sample being studied, no generalization Mean, median, mode Measures of central tendency, measures of dispersion Inferential Used to generalize, apply information from sample to population, includes parametric and nonparametric Confidence intervals, normal distribution Nonparametric: t-test, analysis of variance (ANOVA) Parametric: Chi-square

### Formulation of a Conclusion and Relationship of Results to the Hypothesis

Based on the statistical results of the data analysis, the researcher determines whether the study shows significance. Whether it shows much, little, or no significance, a conclusion can be formulated. From the results discovered, the researcher decides to either accept or reject the null hypothesis of the study. Occasionally, when formulating a conclusion, a researcher may make an error. Errors within research are of two types:

Type I alpha (α) errors occur when, according to statistical results, the researcher rejects the null hypothesis when it is true. The researcher’s conclusion states that a relationship exists when it does not. Most statistical analyses use a level of .05, which means that there is a 1 in 20 chance that a conclusion will state that a difference exists when there is no difference.

Type II beta (β) errors occur when the null hypothesis is accepted but is actually false. The conclusion states that no relationship exists when one actually does (Table 7-5).

Table 7-5

Null Hypothesis

 Null Hypothesis Is Actually … Null Hypothesis Is Accepted Null Hypothesis Is Not Accepted True No error Type I α (alpha) error False Type II β (beta) error No error

## Analysis of the Literature

A thorough review of the literature is often completed to begin development of an adequate plan for collecting the data. Being informed and up to date not only are professional responsibilities but also serve a purpose in each of the roles of the dental hygienist: clinician, educator, advocate, administrator, and researcher.8 The assimilation of information requires more than listening to colleagues and attending occasional Continuing Education programs. An excellent source of information is dental and other scientific literature.

A literature analysis, besides helping the dental hygienist to develop a data collection plan, provides valuable information about theories, methods, and products that are available. Reviewing the literature is an important step in remaining current within the field of dentistry and provides information used to intelligently answer many questions posed by patients. Scientific literature contains information that can help one to maintain competency and helps to set the exceptional practitioner apart from others in the field of dentistry.

Not all dental hygienists receive regular subscriptions to scientific magazines or have access to Internet services. Although these are the most common ways to obtain information about literature, a trip to a local library with a scientific collection can provide essential information. Becoming skillful at obtaining scientific information is not as easy as might be expected. However, it is a skill worth cultivating because it provides a valuable tool for researchers and for practicing dental hygienists.

The topic described next presents an overview of what is available in the form of written resources, how to choose the best sources of this information, and how to critically review this literature. Although a critical analysis of the scientific literature is ideal, it is best to remain open-minded when inquiring about new products, services, and techniques.9

Finally, keeping focused on the research or specific information to be reviewed should help to provide simple yet intelligent answers to the question at hand.

### Selection of Literature

To begin a literature review, the dental hygienist or researcher must select appropriate journals. The scientific writing to be reviewed should be comprehensible to the average dental hygienist who is knowledgeable about the topic area. The selection of literature that is pertinent to the field of dental hygiene will allow the researcher to obtain a complete understanding of the research topic while focusing the research on issues important to dental hygiene. Because of the technicality and intricate scientific detail of its topics, the Journal of Biochemical Research might not be an ideal place to start looking for information on periodontal host factors, whereas the Journal of Periodontology and Journal of Dental Hygiene might be preferable choices. Although both journals publish in-depth scientific literature, the material is tailored to the dental field and thus is relevant and understandable to the average oral health researcher.

Equally important is the selection of a reputable journal. Several aspects lend credibility to a reputable source, including an editorial review board that evaluates each contributed article for accuracy, relevancy of content, and issues involving style and method of scientific writing. This is also known as a “refereed,” or a “peer-reviewed,” journal. Individuals who are considered experts on the article’s content review the articles submitted with a peer review board. A refereed journal ensures that an article is written in appropriate scientific style and that the data published therein reflect current knowledge. A reputable journal is commonly affiliated with a professional group or society, a specialty group, or a reputable scientific publisher. A reputable scientific journal is not a journal that is a popular magazine or published by a commercial firm.

Examples of poor choices for scientific literature include any of the typical newsstand health and recreation journals and glamour and beauty magazines. Professionals should appreciate the fact that although many attractively presented dental publications exist, many are simply glorified advertising brochures and do not represent an acceptable source of scientific material. When selecting a journal article, the reader should be careful to note that the author has the appropriate qualifications. (An attorney writing an article about orthodontics, for example, might not be the most credible source of information.) Authors should also possess experience or a current relationship with the field about which they are writing. If the written work is a research study, there should be evidence of facilities in which to conduct the research and financial support for the project.

Readers usually find a tremendous amount of available information. Although older information may be considered classic and therefore occasionally useful in conducting a review, most often readers want to research the most current information. One classic study was the Vipeholm study, conducted in Sweden in the 1950s.10 The study investigated the incidence of decay in relation to sugar intake and is often mentioned in scientific writing. Although the information from this research has proved valuable, this study was conducted in a disadvantaged population and by today’s standards would not have received approval because of ethical considerations. One of the most often cited concerns about the Vipeholm study was whether the subjects enlisted possessed the mental capacity to give their consent for the research.

Information is usually considered current if it has been published within the last 3 to 5 years. References cited in journal articles should be carefully screened to validate their relevancy and age. Sometimes only a limited amount of information is available on a given topic, and this is reflected in the article. An example is the lack of true research involving herbal or alternative dental therapies.

### Current Topic of Interest Example: Bisphosphonate-Related Osteonecrosis of the Jaw

Dental hygienists working in the community, particularly in association with research institutions such as Veterans Affairs hospitals, often come across topics of interest to the profession affecting patient treatment options to which the community at large has not yet been exposed. Participation in the research related to these topics and publication of the research along with development of subsequent treatment options is often an opportunity for dental hygienists. Review of research is essential in determining not only treatment options but also consideration of patient health outcomes during dental treatment. A current example is the evolution of the diagnosis of bisphosphonate-related osteonecrosis of the jaw (BRONJ). The original research article on the topic appeared in the Journal of Oral and Maxillofacial Surgery in 2004,* and the disease was subsequently detailed in the position paper by the American Association of Oral and Maxillofacial Surgery (AAOMS) in September 2006. The position paper includes oral hygiene recommendations relevant to the practice of dental hygiene in controlling the disease. After further research on the topic within and outside the dental field, recent treatment options were refined and related oral hygiene considerations were continued to be noted as important in the treatment of the disease. A literature search and review shows several current publications available for dental hygienists review and consideration. Based on the new original research recently conducted, the AAOMS updated its position paper in January 2009.

### Evaluation of the Selected Literature

After a journal and the information relevant to the reader’s goals are selected, the reader can pursue a comprehensive evaluation of this literature. Various types of journal articles undergo different types of review. For example, a review article may examine an assortment of studies that have already been conducted and provides an overview of the research that has already been done. A review article can help one formulate an idea or a new research question and can help direct the reader to other sources of information through references cited within the article. Validation of a good review article should follow all of the practices stated earlier, including author expertise, accurate and recent references, and review from a refereed journal.

### The Primary Research Manuscript

Perhaps the most useful type of journal article and truly the archetypal “research paper” is the primary research manuscript (Box 7-4). A primary research study describes the original research, including its methods, materials, results, and conclusions. When one is considering a primary research article, the practice of checking author expertise, accurate and recent references, and peer review from a refereed journal still applies; additionally, other steps are to be followed. The first step is often an assessment of the paper’s abstract.

BOX 7-4   What to Look for in a Research Study Manuscript

A Abstract

• Contains 200 words or less.

• Clearly states purpose of study in the first few sentences.

• Includes brief description of the following: population, type of research, overview of statistics, results, and conclusions.

B Introduction

• Review of the supporting literature.

• Statement of hypothesis (null hypothesis).

• Reason for study.

C Methods and Materials

• Appropriate selection of instrument.

• Appropriate method of conducting research (i.e., prospective, retrospective, randomized, etc.).

• Descriptive enough that reader can replicate the study.

D Results

• Appropriate statistical tests.

• Appropriate display of data.

• Clear and understandable presentation of the data.

• Correct interpretation of data.

E Discussion

• Conclusions are based on fact.

• Results are tied into previous research discussed in introduction.

• Inferences and opinions are stated as such.

• Future plans are included for further research

#### Abstract

A relevant abstract is usually confined to approximately 200 words and concisely defines the study’s purpose, methods, materials, and results. The abstract, a brief description of the research, appears at the beginning of the manuscript and is designed to provide the reader with an overview of the study. Although it may present an idea of what the study involved, the abstract may not always paint an accurate picture of the study and the results. The only way to truly assess a scientific article is to read the content within and critically examine each piece for information.

A primary research article begins with a review of the current literature and an introduction to the study. Within this section, an accurate and complete description of the research problem is given and the purpose of the study is clearly stated. The research question can be stated as a hypothesis and may include objectives to be accomplished.

#### Materials and Methods

The next section of the primary research manuscript, materials and methods, describes the population or sample and the techniques used to gather information about the population studied. One of the primary reasons for disseminating results in peer-reviewed journals (in addition to sharing the results) is so colleagues might duplicate the experiment to validate the results or might modify it in some respects to further refine the conclusions implied in the study. Toward that end, the materials and methods text should be complete enough so that other readers might reasonably expect to recapitulate the experiment and verify the results.

The reviewer will want to determine whether the researcher has selected an appropriate group to test and whether it is one that is relevant to the reviewer’s needs. This necessity for appropriateness applies whether the “group” in question is a population of laboratory animals, human subjects, or tissue culture cells. For example, if reviewers are looking for information regarding nonsurgical periodontal therapy, they should ensure that the study has used a similar population or sample relative to their needs.

The number of subjects in the group is also important. Has the study included enough subjects so that readers can generalize the findings to their group of interest? If the subject is a 35-year-old woman and the research involves men older than 50 years of age, the results may not necessarily be relevant to the reader’s cause; if the subject is a 35-year-old woman but only one or two subjects are involved, the information might also be irrelevant.

Ensuring that no author bias has been introduced is very important. Bias may be defined as any influence that produces a distortion in the results of a study.1 Here are some considerations:

• Is the subject a patient of the researcher, who also happens to be the one who developed the new technique?

• Are all variables in the study controlled for (e.g., diet, standard of living, gender, age, dental history)?

• Is there a control group and an experimental group?

• Is one group receiving the standard treatment and the experimental group receiving the new therapy?

An example of a research study may be the evaluation of the efficacy of sealants in caries prevention. The study design may consist of a group of test subjects (the experimental group) who are to receive sealants and are then to be monitored for a number of years to determine the caries rate. Because reference data are essential, a control group (the group without sealants) is compared with the group with sealants.

Readers must also consider the following issues when evaluating research design:

• If an instrument (e.g., a questionnaire) is to be part of the study, have validity and reliability been previously established?

• Are the conditions under which the treatment is accomplished similar and completely described?

• Are both groups monitored for an adequate period of time to assess long-term results of the therapy?

#### Results

After the methods and materials are described, the results section, including a statistical analysis, follows. The results text should detail how the hypothesis of the study has been tested. Statistical tests should be appropriate for the study and should be described. Tables and graphs may be included to provide a visual representation of the results, but they should be clear and understandable to the reader. The author should justify the statistical method used.

#### Discussion

Finally, a discussion of the conclusions and the inferences drawn from the results of the research is presented. The conclusions are also used to define outcomes of the research.

The conclusion should clearly state the rejection or acceptance of the null hypothesis. It may discuss facts derived from the research but may also include investigator speculation on what the results mean. The conclusion usually discusses the research study’s strengths and weaknesses and may mention further research necessary to obtain the desired results.

Complications observed during the research should also be presented. The results of the study are related to the literature cited. Most important, the conclusions are a direct reflection of the findings. Although speculation may be appropriate, it should be stated as such. It is never appropriate to make statements that are not based on fact or that are not derived from study results.

## Summary

This chapter provides an overview of the basics of research, including the steps in the scientific method, the steps in analyzing the literature, and the components of a primary research manuscript.

Although all research should be conducted according to the scientific method to provide results with a measure of validity and reliability, scientific research remains an inexact science. However, when studies are properly designed and accurately analyzed with the use of the appropriate statistical design, the information obtained not only will be new but also may serve as a springboard for further studies. The inventive and inquisitive practitioner will seek to discover information that enhances the practice of dental hygiene and all its contemporary roles and keeps the profession moving in a forward direction.

1. Formulate a research problem based on a question you have that is related to the field of dentistry.

2. Develop a hypothesis and a null hypothesis for the research problem.

3. Considering your research problem, define the following:

a. Population

b. Sample

c. Experimental group

d. Control group

e. Independent variable

f. Dependent variable

4. Determine whether the data collected from your study will be continuous or discrete. Will nominal, ordinal, interval, or ratio scales of measurement be used?

5. Complete a literature review for the research problem formulated in No. 2, and write an abstract for one of the journal articles examined during your literature review.

6. Using data that you have reviewed or collected, determine the mean, median, and mode.

7. Give five examples of positive and negative correlations. Compare variables related to dental hygiene or to data from articles you have read.

8. Using one of the research studies from your literature review, describe the statistical analysis. Were the statistical techniques used appropriate? Were the data displayed in an appropriate manner?

9. Design and complete a research study or community project following the steps listed in Box 7-1 (the scientific method). Complete your study by creating a poster presentation using appropriate displays of data and the description of your study.

## Dental Hygiene Competencies

Reading the material in this chapter and participating in the activities of Applying Your Knowledge will contribute to the student’s ability to demonstrate the following competencies:

## Core competencies

C.4 Assume responsibility for dental hygiene actions and care based on scientific theories and research as well as the accepted standard of care.

## Community involvement

CM.6 Evaluate the outcomes of community-based programs, and plan for future activities.

## Patient/client care

PC.1 Systematically collect, analyze, and record data on the general, oral, and psychosocial health status of a variety of patients or clients using methods consistent with medicolegal principles.

PC.3 Collaborate with the patient, client, or other health professionals to formulate a comprehensive dental hygiene care plan that is patient-centered and based on scientific evidence.

## Community Case

Allison is a registered dental hygienist who has spent the last 10 years working in a periodontal practice that treats clients referred from several different dental practices in town. Most of the clients present with moderate-to-advanced periodontal disease. Allison has been intrigued by the different information she has seen in several professional journals regarding the link between periodontal disease and heart disease. Allison is also interested in a nutritional supplementation plan she read about that provides protection from inflammation and facilitates wound healing. Allison remembers that much of the information linking periodontal disease and cardiac health indicates inflammatory factors may be a possible culprit. Allison hypothesizes that she can take the relatively small sample her client population provides and make some generalizations regarding oral health as related to cardiac health. After speaking with her employer and gaining regulatory approval for her project, Allison consents and enrolls 100 subjects into her study. Allison enrolls subjects diagnosed with moderate periodontal disease and randomly assigns half into a group treated with standard therapy and the other half into a group treated with standard therapy plus the nutritional supplement purported to provide antiinflammatory and wound-healing benefits. Allison hypothesizes that the group treated with standard periodontal therapy plus the nutritional supplemental will present with a lower incidence of cardiac risk at the end of her study. Allison’s subjects will be followed over the next 5 years. At the end of the study she plans to compare prestudy laboratory and physical analysis with poststudy laboratory and physical analysis to provide information determining whether an increase or decrease in cardiac risk factors occurred in the experimental group. Allison feels she will be able to then generalize her results to other clients in different practice settings who are treated with a similar protocol.

1. The experimental group in this study is which of the following?

a. All the subjects Allison enrolls

b. The subjects receiving standard care

c. The subject receiving standard care plus the nutritional supplement

d. All the subjects Allison attempts to recruit for her study

2. The independent variable in this study is which of the following?

a. The standard periodontal therapy provided to each subject

b. The nutritional supplement

c. The cardiac status of subjects at the end of the study

d. The laboratory tests

3. The data that Allison is collecting to perform her analysis include laboratory parameters (hematology and chemistry). An example of one test result would be a cholesterol level valued at 350. The type of data collected and the scale of measurement most applicable to these data would be which of the following?

a. Continuous and interval

b. Discrete and nominal

c. Continuous and ratio

d. Discrete and ordinal

4. For her data analysis and presentation of results, Allison considers several types of statistical analyses. Allison intends to follow her original plan of applying the information she has collected to patients outside of her study. Which of the following would be the best choice?

a. Descriptive statistics

b. Inferential statistics

c. Nonparametric statistics

d. Power statistics

5. To strengthen her argument and reduce the amount of bias introduced in her study, Allison had planned for some additional data analysis involving laboratory values compared to physical parameters such as age and weight of her subjects. On review of a subset of her data, Allison notices that regardless of the group the subjects were randomized to, as the weight of the subjects increases, a known biomarker for heart disease, the triglyceride level, also increases. When Allison plots out this data she becomes aware of a relationship between her subjects’ weight and triglyceride levels. The information Allison has collected shows which of the following?

a. A positive correlation between the laboratory value and subject weight

b. A negative correlation between the laboratory value and subject weight

c. A positive correlation between the experimental treatment and heart disease

d. Both a and c

## References

1. Polit, D, Hungler, B. Nursing Research Principles and Methods, 5th ed. Philadelphia: JB Lippincott; 1995.

2. Monsen, E. Research: Successful Approaches. Chicago: American Dietetic Association; 1992.

3. Norman, G, Streiner, D. PDQ Statistics. Philadelphia: BC Decker; 1986. [p. 15-18].

4. Cancer Facts and Figures 2004. Cancer Basic Facts: How is cancer staged? Available at www.cancer.org. [Accessed November 2004].

5. Zarkowski, P. Community oral health planning and practice. In Darby M, ed.: Comprehensive Review of Dental Hygiene, 5th ed, St Louis: Mosby, 2002.

6. Rose, L. Overview of biostatistics. In Gluck GM, Morganstein WM, eds.: Jong’s Community Dental Health, 5th ed, St Louis: Mosby, 2003.

7. Potter, P, Perry, A. Research in nursing care. In: Contemporary Nursing: Dimensions and Dynamics. St Louis: Mosby; 1993.

8. Darby, M, Walsh, M. Dental Hygiene Theory and Practice, 2nd ed. St Louis: Saunders; 2003.

9. Hittleman, E, Afes, V. Accessing and reading dental public health research: Evidence-based dental practice. In Gluck GM, Morganstein WM, eds.: Jong’s Community Dental Health, 5th ed, St Louis: Mosby, 2003.

10. Gustafsson, BE, Quensel, CE, Swenander, LL, et al. The Vipeholm Dental Caries Study: The effect of different levels of carbohydrate on caries activity in 436 individuals observed for five years. Acta Odont Scand. 1954; 11:232.

Bibliography

Armstrong, RL. Hypothesis formulation. In: Krampitz SD, Pavlovich N, eds. Readings for Nursing Research. St Louis: Mosby, 1981.

Campbell, JP, Draft, RL, Hulin, CL. What to Study: Generating and Developing Research Questions. Beverly Hills, CA: Sage Publications; 1982.

Huff, D. How to Lie with Statistics. New York: WW Norton; 1954.

Kleinbaum, DG, Kupper, LL, Morganstern, H. Epidemiological Research. Belmont, CA: Lifetime Learning Publications; 1982.

Kraemer, LG. Research and theory development in dental hygiene. In: Darby ML, Walsh MM, eds. Dental Hygiene Theory and Practice. Philadelphia: WB Saunders, 1995.

Norman, GR, Streiner, DL. PDQ Statistics. Philadelphia: BC Decker; 1986.

Pagano, RR. Understanding Statistics in the Behavioral Sciences, 9th ed. Belmont, CA: Wadsworth Publishing; 2008.

Polit, DF, Hungler, BP. Nursing Research: Principles and Methods, 5th ed. Philadelphia: JB Lippincott; 1995.

American University literature review tutorial. www.library.american.edu/Help/tutorials/lit_review/index.html