Random row effect - intercept + coeff ouput #113
-
|
Hi Bert, If my model was a NBinomal with random row row-effect, and I wanted to interpret a couple of significant species-specific coefficients under a given X variable..... is the output a change in counts or a change in the percentage of that species? For example, if species A had an intercept of 0.514, coefX of +0.078, under conditions of X = 30, then Is this an increase of 17.36 % or an increase in 17 individuals at X=30? |
Beta Was this translation helpful? Give feedback.
Replies: 2 comments
-
|
Hello! In my opinion, coefficients in GLM-type models are best interpreted in a relative manner; negative coefficients can be interpreted as decreasing the mean, positive coefficients as increasing the mean. However, if you want a more meaningful interpretation I would suggest looking into resources on GLMs, as there are a lot that will do better in explaining this than I will. Your question relates to log-linear models generally, so models with (for example) a Poisson distribution are similarly interpreted. You could read https://www.middleprofessor.com/post/interpreting-coefficients-in-glms/, which is a resource by Gordana Popovic. Or for example: https://www.r-bloggers.com/2018/10/generalized-linear-models-understanding-the-link-function. |
Beta Was this translation helpful? Give feedback.
-
|
Hi Bert! Thats brilliant! Thanks so much for the help.... I got a little confused looking at sources of GLM interpretation with identity links! |
Beta Was this translation helpful? Give feedback.
Hello!
In my opinion, coefficients in GLM-type models are best interpreted in a relative manner; negative coefficients can be interpreted as decreasing the mean, positive coefficients as increasing the mean. However, if you want a more meaningful interpretation I would suggest looking into resources on GLMs, as there are a lot that will do better in explaining this than I will.
Your question relates to log-linear models generally, so models with (for example) a Poisson distribution are similarly interpreted.
You could read https://www.middleprofessor.com/post/interpreting-coefficients-in-glms/, which is a resource by Gordana Popovic. Or for example: https://www.r-bloggers.com/2018/10/gene…