in QOL scores that could be predicted by this group of predictor variables.
Suppose the researcher finds that emotional distress accounts for a large por-
tion of QOL scores, and employment status accounts for only a small portion
of QOL scores. Using this evidence to determine best practice, nurses should
recognize that interventions directed toward alleviating emotional distress
would have a greater influence on dialysis patients’ perceptions of QOL than
would interventions directed toward altering employment status.
Sometimes researchers want to know how accurately a group of predictor
variables can determine group membership. Determining group membership
is the second aim of predictive correlational designs. For example, this type of
design could be used if a researcher wanted to determine to what degree the
predictor variables of age, self-efficacy, level of education, and level of general
stress predict whether a woman will smoke during pregnancy. Data will allow
the researcher to calculate the odds that a woman will be a smoker versus the
odds that she will be a nonsmoker. The outcome variable is the group mem-
bership because women are either predicted to be in the group of smokers or
in the group of nonsmokers. Although correlational studies are considered to
produce a lower level of evidence, odds ratios are growing in popularity as a
way to inform practice.Model-Testing Correlational Designs
A third type of correlational design is model testing. Researchers use this
type of design to test a hypothesized theoretical model. All related variables
are identified, and specific hypothesized relationships are stated. Researchers
create graphic representations or paths to show the relationships among the
variables. For example, Figure 7-1 is a graphic resulting from a study (Tomake,
Morales-Monks, & Shamaley, 2013) conducted to determine the relationships
among variables related to alcohol problems in college students. Statistical
analysis is done to test all of the relationships at one time. The analysis deter-
mines how well the data collected actually “fit” the hypothesized model. The
better the fit of the model to the analysis, the more likely it is that the predicted
theoretical relationships are true in reality.
Often the term causal modeling is used for model-testing designs, but this is
misleading. Although model-testing designs provide a rigorous test of predicted
relationships among multiple variables based on theory, they can only provide a
suggestion of causality, not true evidence of causality. Model testing and predic-
tive designs, although they establish predictive links between variables, do not
allow researchers to say that variable X causes variable Y. True causality can be
established only with an experimental design. However, because in nursing it is
often unethical to manipulate IVs, model-testing designs provide the strongest
nonexperimental evidence regarding the relationships among variables.KEY TERM
model testing:
Correlational
design to test
a hypothesized
theoretical model;
causal modeling or
path analysis184 CHAPTER 7 Quantitative Designs: Using Numbers to Provide Evidence