PNB 3RM3 Lecture Notes - Lecture 42: Publication Bias, Null Hypothesis, Statistical Inference
Document Summary
Because two groups will always be different from one another: p value is the probability of obtaining a result if the null hypothesis is true** When you do your analysis and your results are not statistically significant: when this occurs, we don"t end up publishing our results and just put it in our. File drawer" stored away, not shared with the world. Because of this problem, we have an overestimation of how large we think effects are even if the true effects are quite small. One alternative to null hypothesis testing is bayesian statistics: alternative way of doing inferential statistics. In bayesian statistics you create a more specific hypothesis. Instead of a traditional p value, can instead calculate probability of a hypothesis given the evidence. Likelihood function multiplied by prior probability over the total probability of evidence. What a stands for: hypothesis of interest. Probability you are calculating is the probability your hypothesis is true given your results.