topic Re: Kolmogorov and Smirnov D in SAS Procedures

Kolmogorov and Smirnov D

S_pera — Mon, 01 Feb 2021 18:28:45 GMT

I find an inconsistency of results with two datasets. With the first dataset, the manually calculated answer is consistent with SAS output. But with the second dataset, it is not the case. Please see the attached Excel sheet. Please let me know where the error is.

Thank you

S_pera

Re: Kolmogorov and Smirnov D

Reeza — Mon, 01 Feb 2021 18:36:28 GMT

Excel file isn't displaying for some reason. You also need to provide the SAS code you used. My first guess is ties - does your data have tied values in it?

Re: Kolmogorov and Smirnov D

S_pera — Mon, 01 Feb 2021 18:57:13 GMT

Hi Reeza

No there were no ties. I am attaching the Excel file again. Please try.

rgds

S_pera

Re: Kolmogorov and Smirnov D

Rick_SAS — Mon, 01 Feb 2021 19:09:30 GMT

Your formula is incorrect. The KS statistic does not use the AbsDiff column.

Instead you need three columns:

1. LowSn = (cumFreq-1)/5

2. DPlus = F(X) - LowSn

3. DMinus = Sn - F(X)

Then KS = max(DPlus, DMinus)

For more information, see What is Kolmogorov's D Statistic and especially the section "Compute the statistic manually".

Re: Kolmogorov and Smirnov D

S_pera — Tue, 02 Feb 2021 05:30:26 GMT

Dear Rick_SAS

Sorted. Thank you very much.

S_pera