The meaning of P-values in statistics is popular

first, the two concepts of "statistically significant" and "significant difference" are explained:

"Statistically significant" and "significant difference" are two different concepts, "significant differences" easily misleading, The original two concept in statistics is often a bit generic, and now explicitly only with "statistically significant". p<0.05 is the assumption that the probability of H0 (i.e. no difference between the two populations) is less than 5%, A is the probability that a Ⅰ class error (rejecting the correct null hypothesis H0) is allowed, Usually set before the hypothesis test, If a=0.05, the probability of being allowed to commit a Ⅰ class error is 0.05, so when p<0.05, Note that, within the scope permitted by a=0.05, the two populations are considered to be different, That is, the difference between the two populations is statistically significant (referring to the statistical parameters of a=0.05); If the p=0.04 at this time, and the first set a=0.01, it is considered that the two general differences are not statistically significant (in the case of statistical parameters of a=0.01), although the two populations did not change, the two general differences did not change; So "statistically significant" is not the same as "significant difference", for example: two groups of numbers: Group A: 3, 3.05, 3.01, 3.04, 2.95; Group B: 3.2, 3.1, 3.15, 3.14, 3.12; The difference between the two groups of numbers (mean) is not large, but the p<0.001, set a=0.01 or 0.05, is considered to be statistically significant in two general differences. This is mainly related to the standard deviation of the two sets of numbers. If the two overall difference is significant, it is easy to think that the two groups of numbers (mean) differ greatly. first-class errors and second-class errors popular explanations： H0: A boy who really loves you H1: A boy who doesn't really love you If H0 is actually set up, and you reject H0 by experience, that is, you reject a boy who you think does not love you and actually love you, then you make the first class of Ⅰ mistake; If H0 does not actually hold, and you accept H0, the same thing, you accept a boy who feels that you love you and does not love you, then you make the first Ⅱ category error. If you want to reduce the probability of committing the Ⅰ and class Ⅱ errors at the same time, you can only increase the number of love N, such as a girl who has experienced n=100 love, the 101th time Love guilty of the first Ⅰ class error and the probability of the first Ⅱ class is much smaller. Statistically, the conservative, traditional view as the original hypothesis H0, novel, interested, want to go to the argument as an alternative hypothesis H1 comparison between P-value and significance level of statistics： It's like a criminal suspect can only assume his innocence before he has no conclusive evidence. Because a man's innocence is more serious than a guilty verdict, and nobody wants to be wronged. So spread out you want to prove that a class of results than the second class good original hypothesis is set to the same class two grades, Among them, there are differences in individual results, which are caused by sampling errors and are purely accidental; The alternative hypothesis is set to one class than the second class grade, in which the sample of the first class of second class grade difference is not accidental occurrence, is highly statistically significant, Therefore, the significance level is generally set to 0.05, when the P-value is less than 0.05, we think that because of the chance of the difference in the probability of the result is relatively small, Therefore, reject the original hypothesis, you can accept a class score better than Class two facts; If the P-value is greater than 0.05, there is not enough evidence to prove that the first class is better than the second class, the probability of the difference in the original hypothesis due to the sampling error is higher, Refuse to accept the assumption that the alternative hypothesis is conservative.