Predictive Modeling, Variable Screening


Another way to screen compares two measures of correlation, Spearman and Hoeffding.



Introduction



The rank option in PROC CORR is handy for examining a large number of variables.



Restructuring the data sets produced with ODS Output



A graph of the Spearman Ranks vs the Hoeffding Ranks is often helpful.



A closer look at the variable ddabal begins with a logit plot.



Use mean substitution for those who do not have a checking account.



Sometimes, writing code that writes code is helpful.



A final check on the created variable, b_ddabal



The slides used in the videos are found here