Timeline for Two-tailed tests... I'm just not convinced. What's the point?
Current License: CC BY-SA 4.0
5 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| May 24, 2018 at 15:13 | comment | added | Flounderer | @FromTheAshes there is nothing wrong with trying to challenge the very foundations. Statistical hypothesis testing is not useless, but it does contain massive logical flaws, and it is absolutely reasonable to challenge them! | |
| May 23, 2018 at 12:18 | comment | added | dbwilson | If I ask you to call heads or tails for a coin toss, the probability of your predicting the outcome is 50/50 (assuming a balanced coin and an honest flipper). However, if I flip the coin first and let you look at it and then make your prediction, it will no longer be 50/50. If you are conducting a one-tailed test with an alpha level of .01 but then flip the direction of the test after seeing the results because p<.01 in the other direction, your risk of a Type I error is no long .01 but much higher. Note that the observed p-value and Type I error rate are not the same thing. | |
| May 23, 2018 at 11:28 | comment | added | FromTheAshes | Say your buddy, Joe, invents a new product that he claims greatly enhances plant growth. Intrigued, you devise a robust study with a control group, and treatment group. Your null hyp. is that there will be no change in growth, your alternative hyp. is that Joe's magic spray significantly increases growth - so a one-tailed test. 2 weeks later, you make your final observations and analyse the results. The mean growth of the treatment group turns out to be over 5 standard errors BELOW the control's. How is this very significant finding any less obvious or valid because of your choice of test? | |
| May 23, 2018 at 11:09 | comment | added | FromTheAshes | To be clear, I am not trying to challenge the very foundations of statistical inference. As I stated, I have only just learned the very basics and am having trouble understanding how any potential findings could be missed by failing to use the correct test. | |
| May 23, 2018 at 9:26 | history | answered | SmallChess | CC BY-SA 4.0 |