Important: Read this before posting to this forum

  1. This forum is for questions related to the use of Apollo. We will answer some general choice modelling questions too, where appropriate, and time permitting. We cannot answer questions about how to estimate choice models with other software packages.
  2. There is a very detailed manual for Apollo available at http://www.ApolloChoiceModelling.com/manual.html. This contains detailed descriptions of the various Apollo functions, and numerous examples are available at http://www.ApolloChoiceModelling.com/examples.html. In addition, help files are available for all functions, using e.g. ?apollo_mnl
  3. Before asking a question on the forum, users are kindly requested to follow these steps:
    1. Check that the same issue has not already been addressed in the forum - there is a search tool.
    2. Ensure that the correct syntax has been used. For any function, detailed instructions are available directly in Apollo, e.g. by using ?apollo_mnl for apollo_mnl
    3. Check the frequently asked questions section on the Apollo website, which discusses some common issues/failures. Please see http://www.apollochoicemodelling.com/faq.html
    4. Make sure that R is using the latest official release of Apollo.
  4. If the above steps do not resolve the issue, then users should follow these steps when posting a question:
    1. provide full details on the issue, including the entire code and output, including any error messages
    2. posts will not immediately appear on the forum, but will be checked by a moderator first. This may take a day or two at busy times. There is no need to submit the post multiple times.

interpretation of class allocation in LC models

Ask general questions about model specification and estimation that are not Apollo specific but relevant to Apollo users.
Post Reply
maa033
Posts: 35
Joined: 23 Jul 2020, 14:00

interpretation of class allocation in LC models

Post by maa033 »

Dear prof Hess & Dr Palma

We have estimated a LC model including interactions for sex and age in the class-allocation part of the model. The model runs well, but having the results we are somewhat uncertain about how to interpret the estimated parameters (gamma_sex, gamma_age)? Are these parameters to be treated like the deltas, and be transformed into class-allocations? Or can the estimated parameters in the model output (see below) be interpreted directly? In the latter case, how do we interpret positive parameters that are above 1, between 0-1, and negative parameters? If they have to be transformed, could you inform us where to find the code for the correct transformation?

Thank you for any help in this matter.

BR, Margrethe

Here is the full LC-model. Class B is used for normalisation:

Model run by snf52282 using Apollo 0.2.8 on R 4.2.0 for Windows.
www.ApolloChoiceModelling.com

Model name : LC_mining_6c_socio
Model description : Simple LC model on mining data with 6 classes - preference space
Model run at : 2023-02-24 16:16:25
Estimation method : bfgs
Model diagnosis : successful convergence
Number of individuals : 874
Number of rows in database : 6992
Number of modelled outcomes : 48944
1 : 6992
2 : 6992
3 : 6992
4 : 6992
5 : 6992
6 : 6992
model : 6992

Number of cores used : 3
Model without mixing

LL(start) : -7855.36
LL (whole model) at equal shares, LL(0) : -7681.5
LL (whole model) at observed shares, LL(C) : -7624.37
LL(final, whole model) : -4975.16
Rho-squared vs equal shares : 0.3523
Adj.Rho-squared vs equal shares : 0.3466
Rho-squared vs observed shares : 0.3475
Adj.Rho-squared vs observed shares : 0.3417
AIC : 10038.33
BIC : 10339.84

LL(0,component_1) : -7681.5
LL(final,component_1) : -16847.53
LL(0,component_2) : -7681.5
LL(final,component_2) : -18703.26
LL(0,component_3) : -7681.5
LL(final,component_3) : -18865.03
LL(0,component_4) : -7681.5
LL(final,component_4) : -10004.35
LL(0,component_5) : -7681.5
LL(final,component_5) : -8657.48
LL(0,component_6) : -7681.5
LL(final,component_6) : -37913.01

Estimated parameters : 44
Time taken (hh:mm:ss) : 00:09:39.36
pre-estimation : 00:00:26.48
estimation : 00:05:20.05
post-estimation : 00:03:52.84
Iterations : 121
Min abs eigenvalue of Hessian : 0.510339

Unconstrained optimisation.

These outputs have had the scaling used in estimation applied to them.
Estimates:
Estimate s.e. t.rat.(0) Rob.s.e. Rob.t.rat.(0)
asc_a -5.462285 0.872569 -6.2600 0.912629 -5.98522
asc_b 4.114880 0.819783 5.0195 0.849996 4.84106
asc_c -3.187890 1.290441 -2.4704 1.573985 -2.02536
asc_d -2.012112 0.261743 -7.6874 0.495902 -4.05748
asc_e -0.681651 0.358088 -1.9036 0.473365 -1.44001
asc_f 0.000000 NA NA NA NA
seabed_a -0.143539 0.044055 -3.2582 0.054137 -2.65143
seabed_b -0.031247 0.036740 -0.8505 0.032915 -0.94935
seabed_c -0.181305 0.023772 -7.6267 0.037356 -4.85349
seabed_d -0.029162 0.007426 -3.9268 0.007238 -4.02885
seabed_e 0.005944 0.016016 0.3711 0.015766 0.37702
seabed_f -1.173473 0.639191 -1.8359 0.899571 -1.30448
salmon_a -0.051310 0.051338 -0.9995 0.041933 -1.22362
salmon_b -0.049509 0.071756 -0.6900 0.079679 -0.62135
salmon_c 0.043669 0.036177 1.2071 0.031168 1.40109
salmon_d -0.070358 0.014830 -4.7443 0.017158 -4.10064
salmon_e -0.021930 0.029689 -0.7387 0.028090 -0.78072
salmon_f -0.370813 0.099677 -3.7201 0.139316 -2.66166
job_a 0.073576 0.026008 2.8290 0.034676 2.12183
job_b 0.003472 0.015182 0.2287 0.015827 0.21936
job_c 0.044455 0.012302 3.6137 0.011965 3.71556
job_d -0.007769 0.003965 -1.9595 0.004399 -1.76621
job_e 0.008962 0.006977 1.2844 0.006823 1.31336
job_f 0.329620 0.155772 2.1160 0.213414 1.54451
cost_a -0.001570 2.9428e-04 -5.3337 3.0156e-04 -5.20491
cost_b -1.5508e-04 2.3052e-04 -0.6727 2.7289e-04 -0.56827
cost_c 6.2605e-04 1.5364e-04 4.0747 1.3058e-04 4.79442
cost_d -3.3901e-04 6.519e-05 -5.2007 1.0014e-04 -3.38536
cost_e -6.4725e-04 1.4222e-04 -4.5511 2.8430e-04 -2.27660
cost_f -0.007151 0.003452 -2.0714 0.004942 -1.44699
delta_a -0.649325 0.615022 -1.0558 0.619977 -1.04734
gamma_sex_a 1.095391 0.276718 3.9585 0.290902 3.76549
gamma_age_a -0.025720 0.008405 -3.0600 0.008696 -2.95771
delta_b 0.000000 NA NA NA NA
gamma_sex_b 0.000000 NA NA NA NA
gamma_age_b 0.000000 NA NA NA NA
delta_c -1.721741 0.627135 -2.7454 0.660136 -2.60816
gamma_sex_c 1.267914 0.269938 4.6971 0.283918 4.46578
gamma_age_c -0.006909 0.007907 -0.8738 0.007800 -0.88568
delta_d -0.333364 0.484720 -0.6877 0.484250 -0.68841
gamma_sex_d 1.162662 0.221214 5.2558 0.223418 5.20398
gamma_age_d -0.014913 0.006471 -2.3045 0.006467 -2.30606
delta_e -1.143919 0.652040 -1.7544 0.672726 -1.70043
gamma_sex_e 0.734089 0.288348 2.5458 0.316989 2.31582
gamma_age_e -0.004824 0.008482 -0.5687 0.009210 -0.52373
delta_f 1.378297 1.276650 1.0796 1.766749 0.78013
gamma_sex_f -0.075044 0.628020 -0.1195 0.809995 -0.09265
gamma_age_f -0.069287 0.022134 -3.1303 0.031132 -2.22562


Summary of class allocation for model component :
Mean prob.
Class_1 0.13370
Class_2 0.20865
Class_3 0.14939
Class_4 0.34085
Class_5 0.13534
Class_6 0.03206
stephanehess
Site Admin
Posts: 974
Joined: 24 Apr 2020, 16:29

Re: interpretation of class allocation in LC models

Post by stephanehess »

Margrethe

this is really more a general choice modelling question than an Apollo question, and you can find details in papers on latent class, or even in my book chapter at http://www.stephanehess.me.uk/papers/bo ... s_2014.pdf

The gamma parameters are part of your class allocation model. For example, the fact that gamma_sex_a is positive means that for someone with that covariate, the likelihood of class a is increased relative to your base class, which is class b (where the gammas are fixed to zero)

Hope this helps

Stephane
--------------------------------
Stephane Hess
www.stephanehess.me.uk
maa033
Posts: 35
Joined: 23 Jul 2020, 14:00

Re: interpretation of class allocation in LC models

Post by maa033 »

Yes, very helpful indeed!
Thank you.
will try to keep posts to the Forum strictly apollo-related!!
I am very happy that this forum exists.

BR,
Margrethe
Post Reply