topic Re: PROC RANK NORMAL=BLOM in Statistical Procedures

PROC RANK NORMAL=BLOM

DataDev2020 — Thu, 02 Sep 2021 08:45:59 GMT

Hi,

The following code would fx. result in af normalised value of 1.8414202783 for original values of 1,000,000.

proc rank data=input_csv normal=BLOM out=output_ranks;
var DEBT_EBITDA_rec;
ranks rank1;
run;

However when calculating the normalised values manually (used in different setup) I end up with a value of 1.734508.

In SAS my proces would be something like:

#Calculate simple ranks with TIES=MEAN -> results in a rank of 2,885 for the value 1,000,000

proc rank data=input_csv out=simpleRankTest;
var DEBT_EBITDA_REC;
ranks rank1;

#Total non-negative observations are 3,009, meaning that I should be able to use the formula (from SAS documentation):

probit(2,885-(3/8)/3,009-(1/4)) = probit(0.95..) = 1.7345..
run;

What do I not see? Thanks

Re: PROC RANK NORMAL=BLOM

FreelanceReinh — Thu, 02 Sep 2021 12:30:27 GMT

Hi @DataDev2020 and welcome to the SAS Support Communities!

You have first calculated the mean rank for each of the 249 observations with DEBT_EBITDA_rec=1000000 and then used these equal values r₁=...=r₂₄₉=2885 in the formula for the Blom normal scores (also note the plus sign in the denominator). But SAS uses the "raw" ranks, i.e., r₁=2761, r₂=2762, ..., r₂₄₉=3009 and then computes the mean of the resulting normal scores.

This is consistent with the documentation of the TIES= option:

MEAN

assigns the mean of the corresponding rank (or mean of the normal scores when NORMAL= is specified)

Re: PROC RANK NORMAL=BLOM

DataDev2020 — Fri, 03 Sep 2021 06:33:03 GMT

I'm embarrassed - much appreciated @FreelanceReinh