Suppose you are comparing the attitudes of men and women to RP. You find an observed difference between the results for two samples (one of men and one of women) - i.e. the sample difference between the two average scores for attitude to RP English is not zero. So clearly the samples are, descriptively, different, but what can you say about the hypothesis about the populations of men and women that you sampled (since it is this "large-scale" hypothesis that you are really interested in)? Common sense says that you could get small differences between samples of men and women without there being any real population difference between men and women, just because samples from populations don't exactly reflect those populations in microcosm. Something called 'sampling error' always comes in. What you want (though you may not realise it!) is to be told a probability: you need to know the probability that you would get a difference the size of your observed one between samples if there were no population difference. If the probability is remote (say 5% or less (p<.05) - the common threshold chosen), then you will conclude that your samples are evidence for a population difference and will say that the difference is, technically, 'significant'. But if the probability is reasonably large (bigger than 5%, p>.05 say), then it is not safe to regard the "no difference" hypothesis as rejectable. The main bit of information you get from any significance test is therefore a probability, which may be referred to as p or sig
ممكن احد يشرحها عالسريع
ايش يقصد ب RP ,P ??