Currently, for each model we have hierarchical, mixed, and separate subtypes. While this is clear in my mind, I think this is probably quite confusing and could be explained better.
Need to convert this into more typical terminology. At the moment I am thinking of specifying these models in terms of:
fixed effects
random effects (currently only by-participant)
hyperpriors.
So we have...
Model name
Fixed Effects
Random Effects (by-participant)
Hyper-priors (thus shrinkage to group mean)
separateLogK
-
logk, alpha, epsilon
-
mixedLogK
-
logk, alpha, epsilon
alpha, epsilon
hierarchicalLogK
-
logk, alpha, epsilon
logk, alpha, epsilon
separateME
-
m, c, alpha, epsilon
-
mixedME
-
m, c, alpha, epsilon
alpha, epsilon
hierarchicalME
-
m, c, alpha, epsilon
m, c, alpha, epsilon (model focussed upon in the paper)
This way of viewing things opens up many exciting possibilities (generative programming)
[x] add this table to the wiki
Refactor JAGS models to make similarities/differences clearer (in prep for auto-generation)
Currently, for each model we have hierarchical, mixed, and separate subtypes. While this is clear in my mind, I think this is probably quite confusing and could be explained better.
Need to convert this into more typical terminology. At the moment I am thinking of specifying these models in terms of:
So we have...
separateLogK
logk
,alpha
,epsilon
mixedLogK
logk
,alpha
,epsilon
alpha
,epsilon
hierarchicalLogK
logk
,alpha
,epsilon
logk
,alpha
,epsilon
separateME
m
,c
,alpha
,epsilon
mixedME
m
,c
,alpha
,epsilon
alpha
,epsilon
hierarchicalME
m
,c
,alpha
,epsilon
m
,c
,alpha
,epsilon
(model focussed upon in the paper)This way of viewing things opens up many exciting possibilities (generative programming)
Refactor JAGS models to make similarities/differences clearer (in prep for auto-generation)