Tag: Proc Corr

SAS: Proc Corr – Different Correlations between variables

A. Pearson Correlation: Between continuous variable and continuous variable.

%macro corr (in, year);
Title "Check Correlation - &in. group";
proc corr data= &in._&year. 
          outp=corr_&in._&year.
          plot=(matrix  scatterplot);
      var exper salary ;
run;
%mend corr;
  • Variables Information
  • Simple Statistics (N, Mean, Std Dev, Sum, Min, Max, Label
  • Pearson Correlations (Correlation Coefficients, Prob)
  • Scatter Plots
  • Scatter Plots Matrix

B. Polyserial Correlation: Between continuous variable and categorical variable.

%macro corr (in, year);
ods output polyserialCorr=corr_polys_&in._&year. ;
Title "Check Polyserial Correlation - &in. group";
proc corr data= &in._&year.  pearson polyserial;
	with gender minority rank;
      var  exper salary;
run;
%mend corr;
  • Variables Information
  • Simple Statistics
  • Person Correlations
  • PolyserialCorr (Wald Test, LR Test)

C. Polychoric Correlation: Between categorical variable and categorical variable. 

%macro corr (in, year);
ods output measures=plcorr_&in._&year. (where=(statistic="Polychoric Correlation"));
Title "Check Polychoric Correlation - &in. group";
proc freq data= &in._&year.;
tables gender*rank minority*rank / plcorr;
run;
%mend corr;
  • Cross-Tabular Freq Table
  • Measures of Association (Polychoric Correlation)

D. Partial Correlation: Pearson correlation between variables while controlling other variables. Following example checks the correlation between salary and gender, controlling for rank and experience.

%macro partial (in, year);
Title "Check Partial Correlation for basesal and gender - &in. group";
proc corr data=&in._&year.  plots=scatter(alpha=.20 .30);
   var salary gender ; /* treat gender as continuous */
   partial rank exper;
 run;
%mend partial;