Page 1 of 1

Error (pi * p : non-conformable arrays) while estimating Hybrid LC with OL

Posted: 25 Jan 2024, 19:44
by mashrur93
Hi,

I am trying to estimate a Hybrid LC model with OL, where the LV is specified in the LC component. The following error is shown:
Error in pi * p : non-conformable arrays when I try adding covariates in the class membership component (CM) along with the class-specific constants (delta) while having LV specified in the latent class component (LC).

I am using apollo 0.3.1.

There are a couple of things that I would like to mention:
  • The model converges without any error when
    • i. LV, along with other covariates (including class-specific constants (delta) and other parameters), are considered in utilities of the class membership component (CM), while no LV but covariates (including ASC and other parameters) are included in the utility specified for the Latent Class component (LC).
    . See attached: HLC with LV in CM only.r
    • ii.LV, along with other covariates (including ASC and other parameters), are considered in utilities of the Latent Class component, while no LV but only class-specific constants (delta) are included in the utility specified for the the class membership model. See attached: HLC with LV in LC only.r
  • regarding 1(ii); however, when I try to add covariates in the class membership model, the abovementioned error occurs. See attached: error_HLC with LV in LC only.r
  • I found a similar error when I tried to add a covariate (g_reg_user) in the CMs (line 110) in the example code, Hybrid_LC_with_OL]. See attached: error_ex_Hybrid_LC_with_OL.r
Link to the codes:
https://mcmasteru365-my.sharepoint.com/ ... w?e=FUElgX

I would really appreciate any help with this. Thanks in advance.

Regards
Mashrur

Re: Error (pi * p : non-conformable arrays) while estimating Hybrid LC with OL

Posted: 22 May 2024, 12:09
by stephanehess
Hi

apologies for the slow reply. We believe we have fixed this issue in the latest development version - can you try it by downloading from http://apollochoicemodelling.com/code.html

Thanks

Stephane & David