Hi, I am having a weird issue: I run a hierarchical model on Stan with the following likelihood: model { for (n in 1:N) { m[n] = alpha[g[n]] + x[n, ] * beta[, g[n]] + effect*w[n]; } uno = rep_vector(1, N); d = sigma2_t*w + sigma2_c*(uno-w); y ~ normal(m, d); } Basically the e…

Can you show a snippet of your design matrix (the thing you’re doing the QR transform on) both before and after you’ve removed the intercept/constant?]

Hi Mike, thanks for all the help on my posts. So this is the matrix of individual-level data without the constant (with the constant I would just add a column on the left [Data Matrix] My hierarchical model has only two levels: individuals and schools (treatment was randomly assigned to schools).…

If I understand it correctly, the intercept had a prior on it before but no longer does. Is that correct? Also, w is a vector of (data) weights on [0,1] which weight two different standard parameters? If those two are correct, I wonder if the absurd school intercepts are those with a w close to 0 o…

[image] simonbrauer: If I understand it correctly, the intercept had a prior on it before but no longer does. Is that correct? Also, w is a vector of (data) weights on [0,1] which weight two different standard parameters? Can you post your full code?

Here is the complete code, let me know if I need to explain some of the context. data { int<lower=0> N; // num individuals int<lower=1> K; // num ind predictors int<lower=1> J; // num groups int<lower=1> L; // num group predictors int<l…

Ah, so this is the model from SUG 1.13 with some slight elaboration. I discovered recently that I’d been misunderstanding that model for years and am still working out what structure it reflects. Any chance you could post your data? That might help me figure out your issue as well as the model struc…

Yes, sure. ordered_groups.csv (40.5 KB) This assigns individual observations to their cluster individual_level_variables.csv (191.0 KB) This are the individual level variables school_level_variables.csv (5.0 KB) This are the school level variables This is the R script to run the model : setwd("…

Ok, so there was a problem with the data which we fixed, now the model behaves better but we only did a small number of iterations (100). Now we are running 1000 iterations with 1000 warmups and see what happens. These are the correct data: individual_level_variables.csv (191.0 KB) school_level_v…

[image] Tommaso11397: whenever you transfer data from Stata to excel to R, things can happen. Glad you found the issue. I’ve had trouble on the backend where Excel likes to interpret coef* (SE) as an equation to be calculated. This is very relatable. You may also want to check out the haven …

Putting prior on the constant separately in a regression

Modeling

jsocolar August 12, 2021, 3:02pm 10

For what it’s worth, you can use Stan more directly from Stata if desired via Stata’s python integration.

Topic		Replies	Views
Sampler getting stuck? Modeling	18	3996	May 26, 2019
Different results using QR parameterization in Hierarchical Model Modeling fitting-issues , cmdstanr	2	564	January 25, 2021
Stan_betareg intercept coefficient unaffected by prior rstanarm	15	2148	August 26, 2017
My first Stan model - hierarchical logistic regression Modeling cognitive-science	28	8719	June 6, 2017
Standardizing predictors and outputs in a hierarchical model Modeling	18	5811	April 30, 2019

Putting prior on the constant separately in a regression

Related topics