| | Re: T-Test - testing for significance
somebody use p<0.05, somebody use even stringent criteria: p<0.01
It is equally saying, there is only 5% (or 1%) of chance that both set of data are coming from the same distribution. Therefore, is P<0.05, I am willing to accept 5% or risk (that I am wrong) to say that both set of data represent different distribution.
For example, 1,2,3,4,5,6,7,8,9,10 is one population, 7,8,9,10,11,....20 is another population, t-test will give you p value of < 0.0002, so it is pretty safe to say that they are different set of data.
However, if you throw a 200 in the second set, you p value is now 0.1245, with is saying you are not so sure about whether they are different anymore.