Timeline for Interpretation of non-significant results as "trends"
Current License: CC BY-SA 4.0
7 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Jul 5, 2019 at 16:23 | history | edited | Sextus Empiricus | CC BY-SA 4.0 |
added 157 characters in body
|
| Jul 5, 2019 at 16:19 | comment | added | Sextus Empiricus | I have adapted the text to take away this misinterpretation. | |
| Jul 5, 2019 at 16:18 | history | edited | Sextus Empiricus | CC BY-SA 4.0 |
added 157 characters in body
|
| Jul 5, 2019 at 16:16 | comment | added | Sextus Empiricus | ....in your case, when you show to challenge the null effect (challenge the idea that one can not predict the coins) by providing a rare case (just like the tea tasting lady) then we should indeed have doubt in the null hypothesis. In practice we would need to set an appropriate p-value for this (since indeed one might challenge the null by mere chance), and I would not use the 1% level. The high probability to doubt the null should not be equated, one-to-one, with the p-value (since that probability is more a Bayesian concept). | |
| Jul 5, 2019 at 16:09 | comment | added | Sextus Empiricus | @David, I completely agree with you that the p-value is more precisely a measure for 'the probability that we make an error conditional that the null hypothesis is true' (or the probability to see such extreme results), and it does not express directly 'the probabilty that the null hypothesis is wrong'. However, I feel that the p-value is not meant to be to be used in this 'official' sense. The p-value is used to express doubt in the null hypothesis, to express that the results indicate an anomaly and anomalies should make us doubt the null.... | |
| Jul 5, 2019 at 8:28 | comment | added | David | "Significant effect just means that you measured the null hypothesis (absence of effect) must be doubted with high probability." I strongly disagree with this statement. What if I told you I can predict the result of any coin flip, we make an experiment, and out of pure luck we get a 1% $p$-value? Would you say there is a high probability of the null hypothesis being false? | |
| Jul 5, 2019 at 8:21 | history | answered | Sextus Empiricus | CC BY-SA 4.0 |