In hypothesis(brmsfit,'Var > 0') output what is the signification of the *, *+, **, or ***

General
,
1

Hello folks,

I am trying to understand what the Stars in the output of hypothesis() represent.

As far as I understand, it is a threshold on the evidence ratio.
If ER>3 we have *
Then I struggle to find some explanation on the other thresholds (*+, **, ***)
An other thing I wonder : Does these thresholds take into account if the hypothesis tested is onesided or two-sided ?
As far as I understand, with a onesided hypothesis (Var <>0), an ER > 19 or <1/19 means 0 being outside the 95% CI, but for a twosided test, the ER (Bayes factor) >3 or <1/3 reveals the same result.
If this is right, do the stars in the hypothesis output reflect this difference in thresholds ?

Exemple of output with different evidence ratio :

image
image1124×324 16.8 KB

As you can see, eventhough some hypothesis tests are onesided (ex: Motor_Delay >0), an ER = 3.6 gives ‘*’.

Thank you for any help

2

I don’t use the hypothesis testing functionality myself, so I could be wrong, but I think brms only outputs a single *, never three *** or *+. Are you sure that the star column in your table is taken directly from brms? Or maybe whoever compiled that table added the extra stars?

I just tried running hypothesis() on a toy example, and it looks like brms prints a comment below the output about how to interpret the star, so hopefully that gives you what you need:

'*': For one-sided hypotheses, the posterior probability exceeds 95%;
for two-sided hypotheses, the value tested against lies outside the 95%-CI.

If you’re actually getting multiple stars and *+ from brms itself can you share some code that reproduces it? I could be wrong about it only using a single star.

3

Hello, thanks for your clarification.
It was indeed a home made script in Rstudio that led to this tables formatting through kable().

Should have known.

Thanks for the clarification

1 Like