is my interpretation of Bayesian ANCOVA correct ?

GuillaumeC

Hello the team,

Thanks again for this marvelous software !

Can you tell me if I am right / wrong before submitting something : )

I want to examine potential differences between groups on a dependent variable (LTPA) while controlling for another variable (AP_GLTQ) through an ANCOVA.

Here the results

Is it correct to say that: " Bayesian ANCOVA indicated that the data were 12 times more likely to be observed under the null hypothesis than the alternative (BF01 = 11.569), which indicates strong evidence in favor of the null hypothesis." ?

Or, as I am interested by the "GROUPE" effect (controlling for AP_GLTQ), should I reported the BF01 = 5.189 ?

Thanks in advance for your help

EJ

Dear GuillaumeC,

Interesting question. It seems that the "controlling" variable really doesn't add much, and including it just incurs a penalty for complexity. Then again, there is an argument for including the controlling variable no matter what (e.g., defining it as "nuisance" and adding it to the null model, in which case the BF = 11.6 obtains). You could also present an "analysis of effects" where you average across all of the models under consideration. I would present the complete reasoning process.

Cheers,
E.J.

GuillaumeC

Sorry I missed your answer !
Thanks a lot EJ, it's clear.