Timeline for p-values and confidence intervals are not matching
Current License: CC BY-SA 4.0
20 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Jan 30, 2023 at 12:05 | vote | accept | Larissa Cury | ||
| Jan 29, 2023 at 22:41 | answer | added | dipetkov | timeline score: 3 | |
| Jan 28, 2023 at 21:40 | comment | added | Larissa Cury | Let us continue this discussion in chat. | |
| Jan 28, 2023 at 21:40 | history | edited | Larissa Cury | CC BY-SA 4.0 |
added 4 characters in body
|
| Jan 28, 2023 at 21:39 | comment | added | COOLSerdash | I'll report the bug. The confidence intervals are calculated with the other function. Report those. | |
| Jan 28, 2023 at 21:38 | comment | added | Larissa Cury |
ps: in addition to that, do you know why confi.int won't work for that? > confint(mod_p, type = "norm") Error in UseMethod("vcov") : no applicable method for 'vcov' applied to an object of class "coef_tbl . Even if I get the right p-values, I'm not getting the CIs and I need to report both
|
|
| Jan 28, 2023 at 21:37 | comment | added | Larissa Cury | Thank you, @COOLSerdash , I tried that, but it didn't work (see the edit above, please). Yes, if course, If I can help others, I'll do that. | |
| Jan 28, 2023 at 21:36 | history | edited | Larissa Cury | CC BY-SA 4.0 |
added 1170 characters in body
|
| Jan 28, 2023 at 21:22 | comment | added | COOLSerdash |
Yes, change YEAR into 0/1 (A = 0, B = 1) and the same with MY_GROUP (G1 = 0, G2 = 1). Just compare the models to see that they're identical. But please report the bug to improve the package! This will help you and others in the future.
|
|
| Jan 28, 2023 at 21:18 | comment | added | Larissa Cury | @COOLSerdash , thank you for the effort, I really appreciate it, I really really need to solve this! You mean changing it into zeros and ones, for example? Essentially, will it still make sense, since the variables are categorical ? (year and group of exam) ? | |
| Jan 28, 2023 at 21:08 | comment | added | COOLSerdash |
I've had a look at the source code of bootstrap_pvals and especially bootstrap_pvals.merMod. I think it's a bug stemming from how the code updates the model. Try converting the group and year variables to binary numeric variables and refit the following model: lmer(CONT_Y ~ year_num + group_num + year_num:group_num + (1|PARTICIPANTS), data = data). Then run the bootstrap again. Now the p-values will be consistent (up to random error). I also strongly suggest submitting a bug report here: github.com/aloy/lmeresampler/issues
|
|
| Jan 28, 2023 at 20:54 | history | edited | Larissa Cury | CC BY-SA 4.0 |
added 108 characters in body
|
| Jan 28, 2023 at 20:49 | history | edited | Larissa Cury | CC BY-SA 4.0 |
added 108 characters in body
|
| Jan 28, 2023 at 20:40 | history | edited | Larissa Cury | CC BY-SA 4.0 |
added 11 characters in body
|
| Jan 28, 2023 at 20:37 | comment | added | Larissa Cury | @COOLSerdash , still, tho, then YEAR2, GROUPB and the intercept should be significant. The last column says that they aren't, tho the CI shows otherwise. Any ideas? | |
| Jan 28, 2023 at 20:35 | history | edited | Larissa Cury | CC BY-SA 4.0 |
added 11 characters in body
|
| Jan 28, 2023 at 20:35 | comment | added | COOLSerdash | Ok, the documentation clearly states that the last column are bootstrap p-values. But as I've said: Both the "traditional" p-value ($0.413$) as well as the bootstrap p-value ($0.568$) are consistent with the confidence interval. | |
| Jan 28, 2023 at 20:33 | comment | added | Larissa Cury |
@COOLSerdash , I honestly don't know. I've never used this bootstrap_pvals() before, I've read the doc, but it didn't quite answer that. However, it was the only function that gave me pvalues for my lmeresampler model
|
|
| Jan 28, 2023 at 20:29 | comment | added | COOLSerdash |
Why are there two columns p-values in the lower table? Please improve the formating of the tables. The value Pr(>|t|) is $0.413$ for the interaction so quite large and not significant at the 5% level.
|
|
| Jan 28, 2023 at 20:26 | history | asked | Larissa Cury | CC BY-SA 4.0 |