0

I have done some analyses in R using GLMs. After fitting the right model to the data, I usually test my model with

anova(model, test="F")

or 

anova(model, test="chi")

depending on the distribution.

For the F-test I get an output with the following 6 columns:

 Df | Deviance | Resid. Df | Resid. Dev |     F  |  Pr(>F)

Is it ok to quote the result of the test like:

F(df, Resid.Df)=F, p= Pr(>F)

If I perform the Chi-sq test on the model, I get an output with the following 5 columns:

 Df | Deviance | Resid. Df | Resid. Dev |  Pr(>Chi)

What do I have to report here? I read that Deviance is corresponding to the Chi-squared value. So is it ok to report something like:

X^2=Deviance, df=df, p=Pr(>Chi)

and is df in this case df or residual df?

And a general question: is it appropriate to report GLM statistics like that?

I would appreciate any help or literature recommendation very much! Thanks!

EDi
  • 13,160
  • 2
  • 48
  • 57
user3187813
  • 1
  • 1
  • 1
  • Conventions for reporting results of statistical tests vary by citation style, and thus by field and journal. I found [this page](http://web.psych.washington.edu/writingcenter/writingguides/pdf/stats.pdf) for APA with a quick google search- I suggest you look around for something similar for whatever citation style you plan to use. – Joe Oct 01 '14 at 14:22
  • Thank you for your reply! Yes, I did a google search and found advices to cite output of different statistical tests. However, I did not find a good practice to cite GLM. And perhaps my major problem is to understand the theory behind the difference of df and resid. df in the output. – user3187813 Oct 01 '14 at 14:28

1 Answers1

0

I aggree with the report like:

X^2=Deviance, df=df, p=Pr(>Chi)

and df=df, not residual df.

X.Chi
  • 45
  • 8