The output of SPSS gives z-score and p-value in any case. But If we have small samples we can't use z-approximation. Can we use the p-value of the output?
AFAIK does SPSS calculate both, the "exact significance" and the "asmyptotic significance" (not sure why it does... but I have seen it in SPSS outputs from Wilcoxon tests), so take the exact value.
But even if you had only the asmyptotic (approximative) p-values: Approximations are bad for small samples. However, if you can "use" such p-values depends on your aims. You surely can use it in a Fisherian way as a "surprise index", but using it for a strict control of error-rates in a decision-context (Neymanian way) won't be a good idea. Here, exact tests would be preferable.
I have just worked my data with critical values and statistical tables (significance when T
I will agree with Konstantinos, use the Monte carlo calculated p-values. If there is a disagreement trust the Monte Carlo p-value.
- Badges
- Science topic