Fisher's Exact Test Not Showing p-value

[schal]

Hi, so I run a Fisher's exact test, and it shows me that the table probability is <0.0001, but doesn't show any other p-values like the one-sided or two-sided p-values

[Reeza]

Show your code and log.

[schal]

PROC FORMAT;
	value $Oxarma '1' = 'None'
				'2' = 'Trace'
				'3' = 'Some'
				'4' = 'Unacceptable';
	value $Oxarmab '1' = 'None/Trace'
					'2' = 'None/Trace'
					'3' = 'Some'
					'4' = 'Unacceptable';
	value $Oxarmac '1' = 'Acceptable'
					'2' = 'Acceptable'
					'3' = 'Acceptable'
					'4' = 'Unacceptable';
	Run;
PROC FREQ data=first;
	tables day*Oxarma / FISHER;
	format Oxarma $Oxarma.;
	Run;
	Quit;

Log:

PROC FREQ data=first;
946 tables cut*Oxarma / FISHER;
947 format Oxarma $Oxarma.;
948 Run;

NOTE: Writing HTML Body file: sashtml.htm
WARNING: Computing exact p-values for this problem may require much time and memory. Press the
system interrupt key to terminate exact computations.
WARNING: There is not enough memory available to compute Fisher's exact test.
NOTE: There were 913 observations read from the data set WORK.FIRST.
NOTE: PROCEDURE FREQ used (Total process time):
real time 5:32.70
cpu time 5:28.39

[SAS_Rob · Accepted solution]

Your answer is in the WARNING.

WARNING: There is not enough memory available to compute Fisher's exact test.

The suggestion would be to either use the asymptotic test or get a Monte-Carlo estimate of the exact p-value by adding the MC option to the EXACT statement.

PROC FREQ DATA=first;

TABLES cut*Oxarma;

EXACT FISHER/MC:

RUN:

[schal]

Monte Carlo Estimate for the Exact Test
Pr <= P 0.9481
99% Lower Conf Limit 0.9424
99% Upper Conf Limit 0.9538

Number of Samples 10000
Initial Seed 358024000


This is the data I get. How reliable is this Monte Carlo estimate? This suggests that the two variables are independent and that the treatment does not affect the outcome, does it not?

[PGStats]

The MC estimate is very reliable. Interpretation in terms of treatment and outcome is left to the researcher. Only he/she knows the details of the data. For instance, why so many data are missing...

[schal]

I see, thank you. I achieved significance for other variable measured. Missing data is due to having sensory panelists that did not participate in every session.

[Reeza]

You should mark SAS_rob's answer as correct, not your own. He provided you the workaround to get your problem solved.

[schal]

Yes, that's what I meant to do, thank you for bringing that to my attention.