Research Study Design

Medical research studies have a number of possible designs. A strong research project closely ties the research questions/hypotheses to the methodology to be used, the variables to be measured or manipulated, and the planned analysis of collected data. CRCL personnel can help you with determining the proper research design and data analyses to adequately address your research your question. CRCL can also assist you in performing the proper statistical analysis on your collected data. Our staff are experts in research methodology and statistical analysis are proficient with multiple statistical methods statistical software packages. The type of research study you conduct determines what the proper data analyses are and what conclusions you can draw from your data.

A Descriptive analysis is provides basic information about a sample drawn from a population of interest. In a descriptive analysis, information is reported about the frequency of and/or percentages of the qualities of interest in the sample (for example the number of men or women with a disorder or the percentage of people for whom a type of cancer progresses into stage 4 after a given amount of time). For variables that measure a quantity measures of central tendency (mean, median, mode) and measures of dispersion (standard deviation, variance, range) may be calculated (for example the average age at which cardiac arrhythmia symptoms first appeared, the median amount of time until a cancer metastasizes, or the modal average person’s rating of chronic pain). Descriptive studies simply describe, they do not inform regarding the relationship between variables nor provide any information about how changes in one variable may cause changes in another.

A correlational study looks to examine a relationship between two or more variables and asks the question: are changes in one variable associated with changes in a second variable? The first variable is often referred to as the independent variable, the predictor variable, or the exogenous variable while the second variable is referred to as the dependent variable, the criterion variable, or the endogenous variable. Examples of correlational studies would be studies that examine the relationship between age and cholesterol level or between dose of Lisinopril and blood pressure. Common statistical methods used in this type of study are Pearson correlation, chi-square, and regression. It is important to remember that correlation does not imply causation, only the existence of a relationship. Why that relationship exists may be due to a causal mechanism but may also be due to other factors which influence both the variables.

Quasi-experimental studies examine the question of whether groups differ but the groups must be naturally occurring groups, not groups created by the researcher. The design is quasi-experimental because of the lack of random assignment of participants (patients, rats, bacterial cultures) into the groups of interest; the groups themselves are predetermined. For example the examination of the relationship between amount of vitamin D in the diet on chemotherapy outcomes for patients with prostate cancer versus colorectal cancer would be a quasi-experimental study because the experimenter has not controlled who has what type of cancer or their intake of Vitamin D. The comparison group may also be a naturally occurring control group, for example an experimenter may study the frequency of cardiac arrhythmias in and elderly population consisting of one group who regularly consume alcohol compared to one group of elderly patients who normally abstain from alcohol. Common statistical analyses in non-experimental studies like these are t-tests, Analysis of Variance (ANOVA), regression, multiple regression, and moderated multiple regression. While non-experimental studies cannot fully prove causation, they can point to the need for a more controlled experimental tests.

The most stringent test of a scientific hypothesis is an experimental study. The hallmark of an experimental design is that rather than simply measuring an independent variable or selecting a preexisting group that differs on that variable, the experimenter manipulates that variable to create experimental and control groups. Random assignment is used to create the groups and, if the sample is large enough, helps to equate the experimental and control groups on all variables except for the variables of interest. An example of an experimental design would be randomly assigning patients with congestive heart failure into one of three groups (two doses of a new beta-blocker or a placebo condition) and examining ejection fraction after three months to determine if heart function differs between the three groups. Common statistical analyses in experimental studies like these are also t-tests, Analysis of Variance (ANOVA), regression, multiple regression, and moderated multiple regression. The stronger the controls in an experimental design the more justified one is in concluding that the manipulation of the independent variable caused the changes in the dependent variable.