Why did p-values ​​become so popular


Next:When and how can Up:What does the result mean Previous:What does the result mean

How can you p- misinterpret values? - Some popular variants

The probability of an empirical date (D.) under the validity of the is not particularly interesting in most cases. It would be far more interesting, based on the knowledge of the result of an experiment, to find the answer to the question about the probability of or to obtain. Other interesting questions would be `` How significant is the effect? ​​'' Or `` How likely is it that I will get a significant result again in a second experiment? '' The p-Wert unfortunately does not provide an answer to any of these questions, but is not infrequently interpreted as if it did (e.g. Tversky & Kahneman, 1971 [66]; Oakes, 1986 [52]). In this way, the probability actually found is often with the inverse probability mistaken. Gigerenzer (1993) [34] calls this `` Bayesian wishful thinking '', since one could calculate such an inverse probability using the Bayes theorem (see Kleiter, 1981 [45], for an introduction to Bayesian statistics). The importance of an effect, the second interesting question, depends primarily on its size, in addition to content-related criteria (see paragraph on effect sizes). And finally, the probability of repeating an experiment can produce a significant result if the same to replicate, can only be estimated if an estimate of the population effect is available beforehand. In addition, the size of the sample must be specified. A p-Value alone does not provide this information.

Usually the substantive or research hypothesis corresponds with the , d. That is, a significant result is interpreted (in different variants) as support for this research hypothesis. But if the research hypothesis is that there is `` no difference '' (e.g. between a control group and an experimental group in front experimental manipulation), or that there is `` no connection '', i.e. if it is related to the corresponds, then particular caution is required. Indeed, in such cases the test strength, the a priori probability of getting a significant result when there is an effect in the population, is often very low; In such cases, however, an insignificant result is often interpreted as a confirmation of the research hypothesis (Sedlmeier & Gigerenzer, 1989 [61]).



Next:When and how can Up:What does the result mean Previous:What does the result mean