0

I'm running a mixed logistic regression on a dichotomous variable in R with glmer, with two fixed effects in interaction: one is a factor with two levels, and the other is a factor with three levels. I wonder if there is any theoretical reason why to use "summary(model)" or "car::Anova(model)" to interpret the output, as I get different results from the two functions.

I copy the code and the outputs of interest here below. Thank you very much, and please excuse me if this question has been already asked! I have not found anything specific on mixed logistic regression.

model <- glmer(DV ~ factor1*factor2 + (1|trial) + (1|participant),
                       data = data, family = binomial)

Anova(model)
OUTPUT:


Response: DV
                     Chisq Df Pr(>Chisq)

factor1            93.4021  2    < 2e-16 ***
factor2            0.3548  1    0.55141

factor1:factor2    5.1397  2    0.07655 .


summary(model)
OUTPUT:


Fixed effects:
                               Estimate Std. Error z value Pr(>|z|)

(Intercept)                     2.4340     0.4137   5.883 4.03e-09 ***
factor1levelB                  -2.0577     0.4190  -4.910 9.09e-07 ***
factor1levelC                  -3.5702     0.4476  -7.976 1.51e-15 ***
factor2levelB                  -0.8429     0.5352  -1.575   0.1153

factor1levelB :factor2levelB    0.4332     0.5467   0.793   0.4281

factor1levelC:factor2levelB     1.2098     0.5683   2.129   0.0333 *
chiaras15
  • 1
  • Hi, welcome to CV! Your results do not seem incompatible to me? If you are referring to the fact that the interaction is not significant in the ANOVA results but one of the interaction comparisons is significant is in the regression summary results, this (probably) is because post-hoc tests have more power as they test a specific comparison instead of making an omnibus test. See here: https://stats.stackexchange.com/questions/9751/do-we-need-a-global-test-before-post-hoc-tests – Sointu Jul 25 '23 at 12:07
  • 1
    Thank you!! I hadn't seen that post and it seems super useful! – chiaras15 Jul 31 '23 at 15:55

0 Answers0