Is there a way to check for correlation between a continuous variable (year of graduation) and a binary variable (yes/no--took a specific course) with n=85, particularly using SPSS? Our hypothesis is that the earlier the graduation the year, the less likely the individual would have had this course. If correlation won't work, can you recommend another statistical test?
$\begingroup$
$\endgroup$
3
-
$\begingroup$ An interesting question here is whether the data will be independent. Of course each datum will be a different student, but if these are students from the same school & who interact, they may be autocorreated. If non-independence isn't an issue, this is a FAQ: you can find answers here, eg. Thus, do you think the students whose graduations are closer in time will be more similar than those further apart after accounting for the possible secular trend? $\endgroup$– gung - Reinstate MonicaCommented Mar 6, 2015 at 21:03
-
$\begingroup$ The run of the mill unpaired t test is, incidentally, a test for association between a (normalishly distributed continuous variable—not sure year of graduation applies—and a binary variable); however the binary variable is typically interpreted as explanatory of the continuous variable, rather than the other way around. But @gung 's concerns still hold. $\endgroup$– AlexisCommented Mar 6, 2015 at 23:05
-
$\begingroup$ You could look at biserial correlation $\endgroup$– mdeweyCommented Mar 12, 2017 at 15:04
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
You should start out with visualization, for instance plotting the zero/one variable (y-axis) against year of graduation. Use jittering to avoid overplotting, and add a simple smooth to easier see any trend. Lowess could be used for smoothing, but there are many possibilities.
A more formal variant is logistic regression of course indicator against year of graduation. For correlation coefficients, see for instance discussion in Correlations between continuous and categorical (nominal) variables. But with your low sample size of $n=85$, do not expect much, and look upon results as descriptive.
answered Oct 19, 2018 at 13:57