ApolloChoiceModelling forum

Dear Stephane,

Yes, I have run the MDCEV model before I try to use MDCNEV, base model, model with socio-demographics, model with socio-demographics and other personal traits, the estimations do not return any errors.

The results from the base model are as follows,

LL(start) : -42920.59
LL(0) : Not applicable
LL(C) : Not applicable
LL(final) : -28564.04
Rho-square (0) : Not applicable
Adj.Rho-square (0) : Not applicable
AIC : 57138.08
BIC : 57168.44

Estimated parameters : 5
Time taken (hh:mm:ss) : 00:00:1.56
pre-estimation : 00:00:0.16
estimation : 00:00:1.14
post-estimation : 00:00:0.25
Iterations : 56
Min abs eigenvalue of Hessian : 2e-06

Unconstrained optimisation.

Estimates:
Estimate s.e. t.rat.(0) Rob.s.e. Rob.t.rat.(0)
alpha_base -18.450 667.38779 -0.02765 0.19994 -92.28
gamma_nrfund 272.595 15.07501 18.08256 12.16243 22.41
gamma_rfund 227.371 9.20559 24.69925 7.87553 28.87
delta_nrfund -9.225 0.03689 -250.04542 0.03974 -232.14
delta_rfund -8.007 0.03289 -243.45575 0.03421 -234.03
sigma 1.000 NA NA NA NA

Thank you as always!

Best,
Yadi

Hi

and what was the LL for the nested one?

Stephane

Hi Stephane,

The LL of MDCNEV model is as follows,

LL(start) : -28564.04
LL(0) : Not applicable
LL(C) : Not applicable
LL(final) : -24889.18
Rho-square (0) : Not applicable
Adj.Rho-square (0) : Not applicable
AIC : 49790.37
BIC : 49826.79

Best,
Yadi

Yadi

okay, so the nested model defintely has a better fit but produces a solution that is outside the accepted range for the nesting parameter. I had another look at your data, and the reason is actually quite obvious. Your two alternatives that are nested together are never chosen together, i.e. they seem to be mutually exclusive. That's not in line with the model

Stephane

Dear Stephane,

Thank you for the reply. Indeed this is the mistake I made when sending out the survey by only allowing the participants to choose one of the funds......
But now problem solved......
Thank you again for your time!

Best,
Lydia