Hi, I tried to model the same question with two different models for an ordinal outcome and ordinal predictor: fit1: crated 3 dummy variables for an 4 levels ordinal variable and each level was always compared to basic level as 0. fit 2: Used the ordinal variable in the model but use mo() Bel…

[image] torkar: Hi Shabnam, You have two models and one is apparently doing a better job in out of sample prediction. Do posterior predictive checks, i.e., pp_check() , to check if the model behaves. You can also simulate data and check if the model recovers all parameters. I used pp_check a…

[image] ssalimi: rec_n_r_bodn2 ~ mo(rec_sevhearing1) + sex1 + age10int1 + yrs2 If I understand the model above correctly then, no. If you want to model each individual I believe you need to use a varying intercept, i.e., (1 | individual). Perhaps @martinmodrak can provide some insights.

[image] torkar: If I understand the model above correctly then, no. If you want to model each individual I believe you need to use a varying intercept, i.e., (1 | individual) . Perhaps @martinmodrak can provide some insights. What I understood about varying intercept is that provides an oveal…

Well, if you have a varying intercept then you will be able to look at ranef(Model) for each individual.

Great. Thanks. I run the model and see how it is and will update here. Thanks again for all.

[image] ssalimi: On a differnt note, can i get individual-based posterior estimate? I mean posterior estimate for each person? … In lme4 it is possible to perform post hoc analyses and store individual prediction value or estimate. Not sure if this is possible with brm or other Bayes models. …

[image] martinmodrak: One method I’ve found useful in some cases is to make predictions for each person while varying only the factor of interest - see PPC for ordered_logistic? for a bit more detail and ideas by others. You got it right. I need individual estimate. The reason for that is I a…

Still @torkar suggestion may work to consider each individual as a level! Just needs more chains as ids are not repeated!!

Do you have only one measurement/person? If that’s the case I believe it would not work.

Yes, in the models above its one measure/person. When we get estimates its mean estimate, I need to get personalized estimate. I will try predict

Results from monotonic effect versus dummy variables in brms

Modeling

alex_hausmann March 13, 2020, 9:46am 25

I did find some older comments of his on this matter (link below). But it was quite some time ago, so maybe one has a more solid idea in 2020?

Topic		Replies	Views
Ordinal response with monotonic predictor brms	31	1775	February 28, 2020
Ordinal Outcome and Ordinal Predictor in brms brms	6	1585	February 22, 2020
Understanding odd estimates from cumulative probit model brms	19	1576	November 8, 2022
Check Interpretation for a logit brms model General interpret-results , brms	12	1824	August 15, 2023
Parameter recover for monotonic effect brms brms	0	394	October 2, 2022

Results from monotonic effect versus dummy variables in brms

Related topics