-1
$\begingroup$

enter image description hereHi I would like to clarify if it's possible to have no significant effects on the glmm model although when you plot the count data it looks like there is a significant difference?

I fitted a GLMM on the following: glmer(Rat.returns~Treatment+(1|Site), data=df, family="binomial")

  • Treatment (control vs road)
  • rat.returns-> was a binary (0=not returned, 1= returned) and the results show no significant effect on treatment on rat returns.

To present my data, I used the means instead but it looks like there is a significant difference. Is it because taking into account the random effects would affect the response variable?

$\endgroup$
  • $\begingroup$ If you could know just by looking at it, there wouldn't be much point to stats, would there? And the whole point of taking into account random (or fixed) effects is that they might explain some of the variation. $\endgroup$ – mkt Sep 16 '16 at 12:00
  • $\begingroup$ It might help to paste in the figure you're referring to & your model output. $\endgroup$ – gung Sep 16 '16 at 12:18
  • $\begingroup$ Hi sorry- just realized the attachment in my initial post failed..Cause I was marked down by an examiner on this - as the examiner said my looking it at a glance there is a significant effect.. $\endgroup$ – minke Sep 16 '16 at 12:49
0
$\begingroup$

I think that part of the confusion might be that there are two variables which might "significantly differ" in the plot. Which variable is it that you think is significantly different in the graph but not the regression? Is it treatment (control vs road) or site (returned vs translocated)? Perhaps you are getting the two of these mixed up. Just from eyeballing the graph it looks to me like there may be a difference in terms of site (comparing green to blue) but not necessarily in terms of treatment (comparing green to green and blue to blue).

$\endgroup$
  • $\begingroup$ My question of interest is if the mean number of rats returning would significantly differ between control & roads . Which means comparing the green bars only. $\endgroup$ – minke Sep 18 '16 at 10:16
  • $\begingroup$ @minke If that is the case, then from what I can tell it looks like there is NOT a significant difference in the graph, so it seems that the graph and model are in agreement. $\endgroup$ – meta-analytics Sep 19 '16 at 22:00

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.