add Cardinality Ratio to output from proc.freq.nlevels;

Ron_MacroMaven — Tue, 02 Jun 2015 16:32:41 GMT

... and JMP, while we are at it.

I have written several papers on Cardinality Ratio and have an sco wiki page on it.

definitions:

* the cardinality of a set is the number of elements in the set

* cardinality of a data set is n-obs

* cardinality of a variable is n-levels

* cardinality ratio is n-levels / n-obs;

* CR is a reducing function, its range is in (0:1]

and it is easier to compare variable's CR than the range of n-levels in (1:n-obs)

if one takes the time to calculate the n-levels of each variable

proc freq data = sashelp.class nlevels;

I know you have the n-obs of the data set,

so, in my book it is a simple step to calculate

cardinality_ratio = n-levels / n-obs;

Why is this a Good Idea?

Because it takes several steps to calculate it.

proc contents ... out = out_contents;

proc freq ... out= out_freq;

proc sort out_contents

proc sort out_freq

data cardinality_ratio;

merge out_contents out_freq;

by name;

cardinality_ratio = nlevels / nobs;

proc summary var cardinality_ratio out = out_means

data cr_type;

select;

when ... cr_type = 'unique';

when ... cr_type = 'many';

otherwise cr_type = 'few';

Having proc freq do the number-crunching makes calculating the

cr-type in (continuous, discrete, unique)

easier.

Re: add Cardinality Ratio to output from proc.freq.nlevels;

Ksharp — Wed, 03 Jun 2015 13:05:07 GMT

What is your question ? you have different version of solution .