topic One-sided Kolmogorov Smirnov test (PROC NPAR1WAY) in Statistical Procedures

One-sided Kolmogorov Smirnov test (PROC NPAR1WAY)

alexfx — Tue, 04 Aug 2020 14:32:29 GMT

I would like to compute one sided kolmogorov-smirnov test using the PROC NPAR1WAY procedure.

Specifically i want to test 2 distributions of real vs. estimated values, but i am not able to distinguish what distribution corresponds to F1 and which corresponds to F2 and i don't understand which criteria SAS uses to determine which values are F1 and F2.

The output is attached.

Thank you!

Re: One-sided Kolmogorov Smirnov test (PROC NPAR1WAY)

ballardw — Tue, 04 Aug 2020 14:46:29 GMT

Please show the code you are using. Best is to post code into a code box opened with either the </> or "running man" icons to preserve formatting. The code is sort of important to know which options were used.

I don't know what you did to make that docx document to attach but it is ugly. You might consider creating an ODS document from SAS instead if you did some form of copy and paste. The result you show is actually sort of hard to read because the columns do not align.

Re: One-sided Kolmogorov Smirnov test (PROC NPAR1WAY)

Rick_SAS — Tue, 04 Aug 2020 15:23:02 GMT

The PROC NPAR1WAY doc has an example similar to what you describe.

Using the EDF option triggers the empirical distribution function tests, which includes the K-S test.

F1 is the distribution of the first class level; F2 is the distribution of the second class level. The first level is shown on the first row of the table of K-S statistics; the second level is on the second row.

Re: One-sided Kolmogorov Smirnov test (PROC NPAR1WAY)

alexfx — Tue, 04 Aug 2020 15:25:31 GMT

ods graphics on;
proc npar1way edf plots=edfplot D data= T1_emp.T1_empresas_gov_teste1;
class chave;
var vol;
OUTPUT OUT= result_teste_gov1;
run;
ods graphics off;

This is the code that i am using.

the results are:

Kolmogorov-Smirnov Two-Sample
Test (Asymptotic)
D = max |F1 - F2|	1.0000
Pr > D	<.0001
 	 
D+ = max (F1 - F2)	1.0000
Pr > D+	<.0001
 	 
D- = max (F2 - F1)	0.0000
Pr > D-	1.0000

Cramer-von Mises Test for Variable vol
Classified by Variable chave
chave	        N	     Summed Deviation
                                     from Mean
Estimado	        10	     0.83750
Observado	10	     0.83750

Cramer-von Mises Statistics
(Asymptotic)
CM	0.083750	       CMa	1.675000

Kuiper Test for Variable vol
Classified by Variable chave
chave	         N	Deviation
                                from Mean
Estimated	10	1.0
Observed 	10	0.0

Kuiper Two-Sample Test (Asymptotic)
K	1.000000	    Ka	2.236068	Pr > Ka	0.0017