I have a 2 x 2 table of binary variables. The sample size is only 21, so I want to get an exact p-value for the likelihood-ratio chi-square. While Fisher's Exact Test provides p values for one-sided tests, there doesn't seem to be an option for getting exact p values for one-sided likelihood ratio chi-square tests. Does anyone know if this is available in PROC FREQ, or do I need to use another procedure?
Here's my code:
Data table;
input n onset hatch;
*where onset=1 for early and hatch=1 for asynchronous;
datalines;
1 1 0
6 1 1
8 0 0
6 0 1
;
proc freq data=table;
weight n;
tables onset*hatch / chisq;
exact lrchi;
run;
and here's the output:
The FREQ Procedure
Statistics for Table of onset by hatch
Likelihood Ratio Chi-Square Test
Chi-Square 3.8191
DF 1
Asymptotic Pr > ChiSq 0.0507
Exact Pr >= ChiSq 0.0873
Fisher's Exact Test
Cell (1,1) Frequency (F) 8
Left-sided Pr <= F 0.9932
Right-sided Pr >= F 0.0783
Table Probability (P) 0.0715
Two-sided Pr <= P 0.1588