Timeline for If the categorical variable is retained in my final model in R, then why does the post hoc analysis say the levels do not differ?
Current License: CC BY-SA 4.0
7 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Oct 21 at 21:48 | comment | added | Sextus Empiricus | Also, small detail in the language use. A test doesn't say whether levels are different. For this you simply fit the result and look at the estimated levels, and compute the difference. What the test does is saying what the 'statistical significance' is of the observed difference. | |
| Oct 21 at 21:46 | comment | added | Sextus Empiricus | @Alan I wouldn't call it common. The two give different results, but are not very different. Typically the conclusion are similar or the same. | |
| Oct 21 at 21:42 | comment | added | Alan | So, anova saying the category is important but the emmeans function saying the levels are not different is a common thing? Thank you for your help! | |
| Oct 15 at 16:35 | comment | added | Sextus Empiricus | We could use the relationship from the previous comment to make adjustments for the likelihood ratio based on curves of p-values for an F-test versus p-values for an LR test. | |
| Oct 15 at 16:29 | comment | added | Sextus Empiricus | Ah, no, they are not the same. There is a relation between the two statistics $$\log \Lambda = -\frac{n}{2} \log \left( \frac{RSS_0}{RSS_1}\right) = \frac{n}{2} \log \left( 1+F \frac{k_1-k_0}{n-k_1}\right)$$ but $\Lambda$ is far from chi-squared distributed if $n$ is small. Replicate t or F test from regression using regression likelihoods | |
| Oct 15 at 16:26 | comment | added | Sextus Empiricus | I wonder whether I used the functions too easily and possibly made a mistake. I would have thought that Anova and LR test are equivalent for the assumption of normal distributed data. | |
| Oct 15 at 16:20 | history | answered | Sextus Empiricus | CC BY-SA 4.0 |