I would appreciate it if anyone can help me with this issue, I have changed my priors, however, it does not work. I try to simulate from a multivariate normal distribution based on a given dataset. The issue is: Chain 1: Rejecting initial value: Chain 1: Log probability evaluates to log(0), i.e.…

Okay, I simulated some data from a random covariance matrix in python and used your model to recover it. (NOTE: I used a zero mean for the random multivariate normal draws). I initialized the above model with inits=0 and it ran fine. Moreover, with N=10000 I recovered my covariance matrix. Can …

That is, change your model to: model<- ' data { int<lower=0> N; vector[15] x[N]; } parameters { vector[15] mus; cholesky_factor_corr[15] chol; vector<lower=0>[15] sigma; } transformed parameters { matrix[15, 15] chol_cov = diag_pre_multiply(sigma, chol); } model { mus ~ normal(0, 0…

Thank you so much for your reply! I also standardized my data, but I got the same error. There are 50 variables and have small variances, and covariances are so close to zero as well. Do you think the problem is the covariance chol ~ lkj_corr_cholesky(1)? model<- ' data { int<lower=0> N; vect…

I noticed the model works for the same datasets but by decreasing the number of variables to 20.

With that information, you could try a prior of chol ~ lkj_corr_cholesky(10) which would indicate a higher belief that the correlation is the identity. Still, it is interesting that you are still getting log(0) did you try initializing with all zeros? (I don’t know how to do this in R, but in cm…

Thanks so much! will do that and let you know the result. Should I change the number of iterations as the dataset has 50 variables?

The number of iterations is fine. Although for testing, you can reduce them down to a few to make sure things are compiling and running Good luck!

Thanks! But unfortunately, got the same result when I tried to sample the whole dataset. SAMPLING FOR MODEL 'anon_model' NOW (CHAIN 1). SAMPLING FOR MODEL 'anon_model' NOW (CHAIN 2). SAMPLING FOR MODEL 'anon_model' NOW (CHAIN 3). Chain 1: Rejecting initial value: Chain 1: Log probability evalu…

Error: Log probability evaluates to log(0), i.e. negative infinity. (Multivariate Gaussian)

Modeling

spinkney December 27, 2021, 11:00pm 11

Try the onion method for that corr matrix Using the onion method in occupancy model hierarchical model - #2 by spinkney

Topic		Replies	Views
Initialisation problems with a multivariate model Modeling	1	390	July 13, 2021
Initialization between (-2, 2) failed after 100 attempts Modeling specification	4	489	May 22, 2023
Initialization Issues with Cholesky Factorization of Covariance Matrix Modeling performance , covariance , prior-choice	1	1051	April 25, 2020
Lkj_corr_lpdf: Correlation matrix is not positive definite General	9	1284	July 12, 2023
A simple multivariate normal distribution failed. Anything wrong? Modeling	3	1030	March 28, 2021

Error: Log probability evaluates to log(0), i.e. negative infinity. (Multivariate Gaussian)

Related topics