This lecture explains the logic of p-values, and how they measure TYPE I error -- the probability of rejecting the null hypothesis of a match between the model and the reality WHEN the null hypothesis is true. NEXT it discusses how multiple tests expand the Rejection Region, which must be considered as a whole to arrive at the p-value. In case of INDEPENDENT test, it is easy to come up with a combined p-value. This is demonstrated with daily exchange rate data for Chinese Yuans versus US Dollars. Various types of MATCH hypotheses are tested in the context of this data. Does the probability remain the same across periods, Is there clustering of up and down returns, and others.
P-Hacking -- getting Significant Results by p-value - The plan was to use a SMALL sample size and monitor SEVERAL INDICATORS, one of them was sure to come out statistically significant. If the plan is HIDDEN, p-value seems significant.
More technical article on p-values - Refers to Ionnadis -- most scientific statistical results are wrong, and other critics of p values
BE L08 Explainng P-values & Adjusting for Multiple Tests - Bayesian Econometrics Lec 8 explains the logic of p-values and also how these are adjusted for multiple tests 1hr16m Video Lecture