Factors coding in GAMLj mixed models

Discuss the jamovi platform, possible improvements, etc.

by eval » Fri Jul 31, 2020 11:11 am

Kudos for the entire team!
I am particularly enjoying Jamovi and I believe this is a great step forward for students and poor/lazy-coders like myself.
I am using the mixed models of the GAMLj package and I really like it.
The only one thing may work best for the analyses I am currently running would be having the opportunity to apply user-defined coding for factors (for example I cannot simply contrast level 3 vs. 2 while contrasting 1 vs. 2 as far as I can understand). Not sure if this is something the developer may find useful.

Thanks for this great piece of software!

Ref
https://stats.idre.ucla.edu/r/library/r-library-contrast-coding-systems-for-categorical-variables/
eval
 
Posts: 2
Joined: Fri Jul 24, 2020 10:37 am

by MAgojam » Sat Aug 01, 2020 1:19 pm

Hi, @eval.
You have already taken a look here:
https://gamlj.github.io/rosetta_contrasts.html

In addition to choosing (Factors Coding) the type of contrast, have you tried to sort the levels of the Variable (DATA-> DATA VARIABLE) differently, compared to how they are by default?
Example: from (1 2 3) to (3 2 1) or (3 1 2) etc.

Cheers,
Maurizio
User avatar
MAgojam
 
Posts: 93
Joined: Thu Jun 08, 2017 2:33 pm
Location: Parma (Italy)

by eval » Sat Aug 01, 2020 4:15 pm

Dear Maurizio,

thanks for replying.
I think I did. Let's assume 1 is my control level.
Almost all the coding options won't run a plain 3 vs.2 without ditching the 2 vs. 1 comparison.
The Helmert will run 1 vs (3, 2) and repeated will compute 2 vs. 3 but not 1 vs. 2.
So eventually the full set of comparisons can only be obtained with the posthocs.
But I may miss something!
eval
 
Posts: 2
Joined: Fri Jul 24, 2020 10:37 am


Return to General