0
$\begingroup$

enter image description here

I have this output from a R ANOVA analysis and the graph attached. I want to verify if the interpretation for the scenario is correct?

                      Estimate   Std. Error z value Pr(>|z|)   
(Intercept)             1.7918     0.6236   2.873  0.00206 **
typeHigh               -1.5911     0.7687  -2.070  0.01847 * 
statusBlack            -1.2528     0.7843  -1.597  0.11018   
typeHigh:statusBlack    2.4384     1.0628   2.294  0.01177 * 
  1. Line 4 of result -- The difference between the status: Black and Gray is different for the type: Low vs High. Here, At low: black decreases w.r.t to gray and at high it increases w.r.t to gray

  2. Line 1 of result -- How do I interpret the first line alone?

  3. Finally, is it correct to say that as we go from low to high: Gray decreases significantly i.e. Gray in "Low" is different from Gray in "High". Also, does black increase significantly i.e. Black in "Low" is different from Black in "High".

$\endgroup$
  • $\begingroup$ I ran a glm negative binomial function, so the values are from the estimate. Ignore those $\endgroup$ – Biotechgeek Jul 24 '19 at 16:43
  • $\begingroup$ negative binomial and ANOVA are totally different things. Then my first comment based on ANOVA is wrong. $\endgroup$ – user158565 Jul 24 '19 at 17:03
1
$\begingroup$

A significant effect in an ANOVA analysis is interpreted as: At least one level of the categorical variable has a mean, which is significantly different from at least one other level. It tells you nothing regarding which group is different from any other group.

When a categorical by categorical interaction is added to the model the groups are broken down into their memberships on both variables. In your case, the 4 groups would be: low and black, low and grey, high and black, & high and grey.

The interaction coefficient simply tells you that one of those 4 groups means is statistically significantly different from at least one other.

To understand which group is different, many methods exist. p-value corrected pairwise tests can be used, dummy coded regression models, or simply graphical methods.

$\endgroup$
  • $\begingroup$ The model is not ANOVA, it is based on negative binomial, as OP said in the comment. $\endgroup$ – user158565 Jul 24 '19 at 17:51
  • $\begingroup$ I see that, but also see that they say to ignore that output. I provide an answer to the question asked, and the 4 points put forth in the OP. $\endgroup$ – Roman_Numerals Jul 24 '19 at 19:20

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.