V II Journal
PUH 5302, Applied Biostatistics 1
Course Learning Outcomes for Unit VII
Upon completion of this unit, students should be able to:
1. Explain the basic concepts of biostatistical analysis.
2. Analyze relevant scientific evidence.
4. Recommend solutions to public health problems using biostatistical methods.
5. Analyze public health information to interpret results of biostatistical analysis. 5.1 Discuss examples of dependent and independent variables in public health. 5.2 Explain the principles of multivariable methods.
6. Summarize the major principles for determining sample size and power.
7. Evaluate the role of biostatistical analysis in public health research.
Course/Unit Learning Outcomes
Learning Activity
1 Unit VII Final Project
2 Unit VII Final Project
4 Unit VII Final Project
5.1 Unit Lesson Chapter 9 Unit VII Final Project
5.2 Unit Lesson Chapter 9 Unit VII Final Project
6 Unit VII Final Project
7 Unit VII Final Project
Reading Assignment
Chapter 9: Multivariable Methods
Unit Lesson
Welcome to Unit VII. In the previous unit, we discussed the major principles for determining sample size and power by computing margin of error, effect size, and variability of the outcome of affect size.
In this unit, we will learn how to analyze public health information and interpret results of biostatistical analysis. In doing this, we will define and provide examples of dependent and independent variables and close with a discussion on multivariable methods.
Many public health research studies, especially those involving socioeconomic and health issues deal with research variables. A variable may be considered anything that can be quantified and varies when exposed to certain conditions or factors. To better understand this concept, let us examine some definitions and then move on to discussing dependent or independent variables.
UNIT VII STUDY GUIDE
Public Health Information: Data and Interpretation
PUH 5302, Applied Biostatistics 2
UNIT x STUDY GUIDE
Title
Definitions
Attributes are the different values included in a variable; for example, males could also include boys.
An exhaustive list is a list that contains all possible answers; gender is an example. Gender can include male and female as well as transgendered male and transgendered female.
Mutually exclusive attributes are attributes that cannot take place at the same time. That means the
concept of all that apply responses cannot be included in a mutually exclusive attribute.
Quantitative data involves numeric, statistical, or mathematical analyses from which the researcher is able to draw inferences.
Qualitative data are subjective and involve the human opinion.
Units are the different classifications of variables (Changing Minds, n.d.).
Types of Variables
There are different types of variables. Some of these are briefly described below. We will then examine dependent and independent variables in more detail.
Descriptive variables are those variables that the research discusses without relating them to anything such as gender or educational experience.
Categorical variables are variables that result from a selection from categories, such as “satisfied,” “highly dissatisfied,” or “dissatisfied.” Categorical variables also include nominal and ordinal variables.
Numeric variables are those presented in numbers, such as weight or sizes. Discrete and continuous variables are numeric variables. In a research study, these variables may either take the place of dependent or independent variables, based on the research question guiding the research and the type of research design (Changing Minds, n.d.).
Dependent and Independent Variables In most research, especially where the researcher is interested in the effect of one variable on the other, it is important to know which variable is the dependent or independent variable. The dependent variable is the variable a researcher is interested in; the researcher plans the study to measure the behavior of the dependent variable (Sullivan, 2018). On the other hand, the independent variable is the variable that effects the behavior of the dependent variable. In many cases, the researcher manipulates the independent variable to cause the dependent variable to change. The researcher has control over the independent variable but not the behavior of the dependent variable. Let’s examine some of the problems that may arise from dealing with dependent and independent variables. Unwanted Influence by Confounding Variables Confounding is the alteration of an effect of a risk factor to a variable (Sullivan, 2018). A confounder is related to the risk factor and to the outcome. Assessing confounding can be done by formal tests of hypotheses and clinically meaningful associations. In other words, sometimes, when studying a dependent variable, results become complicated. For example, you are studying the performance of two groups to know the effect of one group on another, but the two groups are doing about the same. This could be caused by a confounding variable, which is defined as an interference caused by another variable (Sullivan, 2018). Confounding variables are also called extraneous variables. Extraneous variables are variables other than the independent and dependent variable that could strongly influence the results of a study. Let’s consider the example of hungry people competing in a hundred-meter dash. There are several confounding variables we need to consider:
metabolism and weight of the individuals,
age of the competitors,
size of the competitors,
the time of year, and
location of the experiment.
PUH 5302, Applied Biostatistics 3
UNIT x STUDY GUIDE
Title
Effect modification occurs when a different relationship between the risk factor and outcome occurs because of another variable (Sullivan, 2018). The third variable then can aggravate or mask the association between the risk factor and the outcome. In dealing with dependent and independent variables, the researcher must be cognizant of the fact that he or she may deal with multiple variables at a time. Therefore, the researcher must be able to use statistical tests to analyze all the variables at one time in order to examine the effects of the variables on one another. An example of a test could be the correction between the independent and dependent variables with several sub variables. For example, a researcher may want to know the effect of nurse supervisor leadership styles (e.g., situational leadership-telling, selling, participation, and delegating leadership styles) on a subordinate nurse’s job satisfaction using Spector’s 1997 job satisfaction scale. The scale has eight facets including pay, supervision, and co-worker. Both the dependent (job satisfaction) and independent variables (leadership styles) have several sub units or variables. The best analytical method to use here is multivariable analysis, which includes multiple advanced techniques for examining relationships among multiple variables at the same time. Multivariable Analysis Multivariable analysis (MVA) involves observations and analysis of more than one statistical outcome variable at a time (Sullivan, 2018). There are many variables in play when analyzing data using multivariable analysis, and the effect of those variables on the responses of interest is essential to the researcher. Multivariable analysis is of immense importance to researchers in several ways. With the use of factor analysis, a form of multivariable analysis, researchers are able to summarize statistical data from tables. They can identify outstanding patterns in the data, such as groups, outliers, and trends. They can also analyze groups in the table, how these groups differ, and to which group individual table rows belong. This type of analysis is called classification and discriminant analysis. Classification and discriminate analysis helps researchers to find relationships between columns in data tables—for instance, relationships between obesity and age or cardiovascular disease. The objective is to use one set of variables to predict another, and this analysis is called multiple regression. We will discuss this more later in the lesson.
Multivariable statistical analysis is especially important in social science research because most social scientists do not use randomized laboratory experiments. Instead, many social scientists must use groups with clear initial differences that could affect the outcome of the study. Multivariable techniques attempt to statistically account for these differences. Researchers use statistical software programs such as the Statistical Package for the Social Sciences software package (SPSS) to perform multivariable statistical analyses. Types of Multivariable Analysis There are various types of multivariable analyses. The researcher must be cognizant of the type he or she intends to use and how those tests relate to the research questions. The first step in multivariable analysis is the quality of the data. Before any analysis is done, the researcher should be able to prepare data in order to improve the quality of the data before any analysis. Data quality and preparation have been discussed in previous lessons. Let’s look at some types of multivariable analysis. Multiple regression is the most common test used in multivariable analysis. It examines the association between a dependent variable and at least two independent variables. The test involves finding a linear relationship and, hence, there has to be normality and linearity in the data. The test could be used for forecasting (Richarme, 2016). Discriminant analysis helps to group observations into homogeneous groups—groups of the same kind such as smokers and nonsmokers. It is important that the independent variables or data have a high degree of normality to get the best results (Richarme, 2016). Multivariable analysis of variance tests help to examine multiple categories of independent variables and two or more dependent variables. Here, the independent variables are categorical wherein the dependent variables are numerical. It is important to note that too many observations may cause the technique to loose practical significance (Richarme, 2016).
PUH 5302, Applied Biostatistics 4
UNIT x STUDY GUIDE
Title
Factor analysis reduces a large number of variables to a smaller number of factors. Sometimes, there may be too many variables in a research design. For researchers to effectively examine these variables, they group these variables into factors, thus forming a set of independent variables with no dependent variables. There are two main types of factor analyses: common factor and principal component analyses. Common factor analysis is based on variance that is common to the factors while principal factor analysis is based on the total variance of the factors (Richarme, 2016). Cluster analysis reduces a large volume of data into smaller groups with significance that can be analyzed. The main problem with this technique is the problem of outliers—irrelevant data or observations. In order to have a good cluster, there are four rules to observe.
Clusters must be different.
Clusters must be reachable.
Clusters must be measurable.
Clusters must be large enough to create an effect (Richarme, 2016).
Multidimensional scaling (MDS) is done to transform similar judgments or preferences of consumers into distances represented by multidimensional space. This technique helps to group consumers by locations for product distribution or marketing (Richharme, 2016). There are other multivariable methods, but for the purpose of this lesson, we will only discuss the techniques above. If you are interested in learning about other methods, please feel free to independently research this topic. In summary, multivariable analysis is vital for analyzing multiple variables at the same time. Multiple techniques have been advanced by researchers, some of which we have discussed in this lesson. However, there are various other methods involved in multivariable analysis. Researchers and experts in statistical analytics have come up with many statistical software programs that help in performing some of these analyses. A typical example is the SPSS software program.
References Changing Minds. (n.d.). Choosing a sampling method. Retrieved from
http://changingminds.org/explanations/research/sampling/choosing_sampling.htm Richarme, M. (2016). Eleven multivariable analysis techniques: Key tools in your marketing research survival
kit [White paper]. Retrieved from https://www.decisionanalyst.com/whitepapers/multivariate/ Sullivan, L. M. (2018). Essentials of biostatistics in public health (3rd ed.). Burlington, MA: Jones & Bartlett
Learning.