search

pmahon3 / UWO-MOTUS-Analysis

0 stars 0 forks source link

Naming conventions #5

Closed pmahon3 closed 3 years ago

pmahon3 commented 3 years ago

Consider:

delta.prime[i] ~ dnorm( muDelta.prime, 1 / sigmaMuDelta.prime^2 )

and one of it's hyperpriors,

muDelta.prime ~ dnorm( muMuDelta.prime, sigmaMuMuDelta.prime^2)

Would it make more sense to say sigmaDelta.prime is the standard deviation of delta.prime and sigmaMuDelta.prime the standard deviation of muDelta.prime in the sense that the sigma prefix in the name is an operator over the remaining portion of the variable name? Like,

delta.prime[i] ~ dnorm( muDelta.prime, 1 / sigmaDelta.prime^2 )

muDelta.prime ~ dnorm( muMuDelta.prime, sigmaMuDelta.prime^2)

sjbonner commented 3 years ago

Yes, this would absolutely make sense -- but I still find the naming somewhat hard to follow. I'm going to create a new branch in which I can modify the names in the JAGS code and see if I can come up with anything clearer.

sjbonner commented 3 years ago

Hey @pmahon3 . I've updated the file Model.R in the new branch I created. I tried to change the names of the parameters in a way that is consistent and would make it easy to identify what is what. I commited after each change so there should be a trace of exactly what I did, and I also added an explanation in comments at the top.

I did add a little to the model by allowing for the mean to vary by day within individual and period. We should also allow the variance to do the same -- but I need to think about how to model this.

Let me know what you think!

pmahon3 commented 3 years ago

Hey Simon, that looks good to me, the specified indexing really helps as well. I'll close this issue and merge these changes into master. I believe there is one additional area where we could add variation, I'll pick up that conversation on the other issue.