Hi all, I’ve been trying to implement a generalised linear latent variable model (GLLVM, see Warton et al. 2015 ), to analyse multivariate data corresponding to species communities. The model is as follows, for species j at site i: y_{ij} \sim NegBinomial(exp(m_{ij}), \kappa) \\ m_{ij} = a_i + \be…

Of course! Here’s the simple case with no covariates: data { int<lower=1> N; //Number of samples int<lower=1> S; //Number of species int<lower=1> D; //Number of latent dimensions array[N, S] int Y; //Species matrix } transformed data{ // Number of non-zero lower triangular factor loading…

Identifiability of GLLVM (factor analytic model)

Modeling

andrjohns October 28, 2021, 11:46pm 4

For dealing with reflection invariance like this, I’ve also had a good deal of success in the past using the sign-correction approach from this post: Non-convergence of latent variable model - #13 by Mauricio_Garnier-Villarre

2 Likes

Topic		Replies	Views
Marginal predictions and cross-validation with latent variable model Modeling loo , ecology , marginal-likelihood , cross-validation	17	2090	April 27, 2022
Multiple modes in latent variable model -- identifiability, or something else? Modeling	0	488	May 16, 2018
Help with factor analysis (latent variable model) Modeling specification	22	5213	June 11, 2018
Multivariate latent variable model (factor analysis) in brms brms brms	3	778	November 28, 2023
Constructing factor loadings as a sum of two parts Modeling ecology	10	670	May 16, 2022

Identifiability of GLLVM (factor analytic model)

Related topics