MNL analysis does not show SE and p-value

This forum is for posts covering broader stated choice experimental design issues.

Moderators: Andrew Collins, Michiel Bliemer, johnr

MNL analysis does not show SE and p-value

Postby apk0022 » Thu Oct 24, 2024 2:26 am

Hi,
I am very new to DCE study.
I launched the pilot study and tried to analyze the pilot data using MNL in R to get the priors.
I got the estimates' values but did not get the SE and p-values to determine the significance of the coefficients.

Here is what the console says about optimization diagnosis:
Model diagnosis : Relative function convergence
Optimisation diagnosis : Inconclusive test
hessian properties : Some eigenvalues are zero
maximum eigenvalue : 0
reciprocal of condition number : not calculated (Hessian is not negative definite)
Number of individuals : 30
Number of rows in database : 330
Number of modelled outcomes : 330

Number of cores used : 1
Model without mixing
LL(start) : -228.74
LL at equal shares, LL(0) : -228.74
LL at observed shares, LL(C) : -227.18
LL(final) : -191.77
Rho-squared vs equal shares : 0.1616
Adj.Rho-squared vs equal shares : 0.1092
Rho-squared vs observed shares : 0.1559
Adj.Rho-squared vs observed shares : 0.1075
AIC : 407.53
BIC : 453.12

I could not figure out the exact issue and am seeking suggestions on how to resolve this issue.

Thank you in advance.
apk0022
 
Posts: 4
Joined: Sat Oct 05, 2024 2:51 am

Re: MNL analysis does not show SE and p-value

Postby Michiel Bliemer » Thu Oct 24, 2024 6:55 am

There could be various issues. You could have specified an unidentified model, for example by including too many constants. Or your data has multicollinearity (check your experimental design). Or you are trying to estimate too many parameters with too little data.
An eigenvalue of zero typically indicates that at least one parameter has no information and hence it cannot be estimated and the covariance matrix cannot be calculated. You will need to identify which parameter that is, often the one that has a weird value (extremely large or small).

I assume that you are using Apollo in R to estimate the model? If so, please post your question on the Apollo forum:
http://www.apollochoicemodelling.com/forum/

Michiel
Michiel Bliemer
 
Posts: 1879
Joined: Tue Mar 31, 2009 4:13 pm

Re: MNL analysis does not show SE and p-value

Postby apk0022 » Thu Oct 31, 2024 6:12 am

Thank you so much for your response.
I missed to specify the reference level for the constant it seems.
I got the values after I fixed that.

Asmita
apk0022
 
Posts: 4
Joined: Sat Oct 05, 2024 2:51 am


Return to Choice experiments - general

Who is online

Users browsing this forum: No registered users and 6 guests