Tests of association are
based on the correlation among two or more variables. A simple
(zero-order) correlation can vary between -1 and +1 with a correlation of 0
meaning there is no relationship among two variables. A negative
correlation means that as values on one variable rise, those on the other fall.
With a positive correlation, increases on one variable are associated with
increases on the other variable. Usually, measurement error results
in correlations that are imperfect (i.e., not -1 or +1), but
correlations of 0.8 or above, can be considered high. Correlations of 0.2
are relatively weak, those of 0.4-0.6 are moderate.
When squared, a
correlation tells you how much of the variability in the two variables is
shared. With a correlation of 1.0, there is perfect overlap. With a
correlation of 0.6, 36% of the variance is shared (overlaps). The higher
the correlation, the more that knowing the score on one variable tells you about
scores on the other variable.
Do not mistake this for
causation. When one variable "causes" another, then they will be highly
correlated. However, variables can correlate for many reasons, sometimes
because a third variable is operating. And, even when highly correlated,
we do not know the direction of the relationship. Simple correlations give
us important information, but more complex research designs are needed to imply
There are many extensions
of the correlational approach that are used in research. One common
approach is referred to as regression analysis. In this approach, the
correlations among a set of predictor variables and an outcome variable can be
computed. For example, we can look at the independent contribution of a
series of variables (e.g., time with the company, satisfaction with compensation
and benefits, ratings of "my manager", etc.) to an outcome variable (e.g.,
"employee satisfaction"). This approach can help us to better understand
the relationship among different variables and give us some indication of which
ones are more strongly related to our desired outcomes.
This type of approach can
be further extended to develop highly sophisticated procedures for statistical
modeling. Statistical modeling can help us to understand the complex
interplay that underlies business relationships. Click on the link below
to learn more about statistical modeling.