Proc Logistic or Proc Corr?

[hpatel3]

Hi Guys,

I've posted a bit on this topic, so sorry if you've seen this before but I figured this might be a better board to get my specific question asked.

I am honestly fairly new to SAS and was thrown in the deep end, so a bunch of the technical jargon goes over my head, FYI.

I'm trying to run a Multivariable Analysis that wants to see if there is a correlation between status of disease (1= they have it 0= they don't) and 7 risk factors (symptoms).

5 of them  are categorical (1= yes 0-no) 

2 of them are continuous 

I've dug it down to the fact that  I need to use Proc Corr ( I think)

but do I use Kendall's Tau as my R2 and P value or Pearson's for my my R2 and P-value? ..or both depending on whether it's continuous or categorical?

----

But HONESTLY, wouldn't it be easier to use proc logistic since it doesn't worry about categorical or continuous and the outcome is binary?

[pink_poodle]

It probably is easier to use Proc Logistic.

You can calculate the point-biserial correlation coefficient for 0/1 coded categorical variable with two categories using the Pearson's method. 

With respect to continuous variables, 

it depends on whether your data is 1. normally distributed and 2. likely to have outliers. The following thesis showed that Pearson's can be used for non-normally distributed data, but, because it is sensitive to outliers, Pearson's is not good for datasets where outliers are very likely:

Pearson’s, Spearman’s and Kendall’s correlation coefficients are the most commonly used measures of monotone association, with the latter two usually suggestedfor non-normally distributed data. 

Pearson’s correlation coefficient could offer a substantial improvement in statistical power even for distributions with moderate skewness...

Nonetheless, because of its known sensitivity to outliers, Pearson’s correlation leads to a less powerful statistical test for distributions with extreme skewness or excess of kurtosis (where the datasets with outliers are more likely).

In conclusion,... Pearson’s correlation coefficient could have significant advantages for continuous non-normal data which does not have obvious outliers.

Reference

Pearson's vs. Kendall's Correlation for Continuous Data

[PaigeMiller]

@hpatel3

Why do you start a new thread for essentially the exact same question? That's not necessary, it's not good, and people get confused by trying to assimilate replies in two different threads. Why don't we continue this at your original thread which is: https://communities.sas.com/t5/SAS-Programming/Multivariate-analysis-Do-I-use-Proc-Corr-or-Proc-GLM-...

[hpatel3]

Will do, thanks!