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.

Fully correlated MXL model

Ask questions about model specifications. Ideally include a mathematical explanation of your proposed model.
Post Reply
nicok
Posts: 1
Joined: 03 Jun 2021, 06:25

Fully correlated MXL model

Post by nicok »

Hi!

First of all, thanks a lot for having developed such as great package! It is very useful and likely a game changer - so congratulations!

I would like to know whether this is possible to simplify the coding/specification of the random terms in apollo_randCoeff().
I've tried different things and it seems that the different random coefficients need to be manually specified (eg mu_x1 + sd_x1 * draw_x1).
This is fine when the model has a limited number of parameters but it can quickly become tedious and error prone when model has many more random parameters. I initially tried to use a loop (see below), as this can be done for the specification of the indirect utility functions, but it does not seem to work.

I am especially interested in estimating a fully correlated MXL model. I've initially used the bayesian approach to "estimate" such model but there is something odd with the results (ie, the improvement in the loglik is way too large to be true) and when i use the bayesian approach to estimate an uncorrelated (diagonal) MXL model, the results don't match what i obtain with the "gmnl" (mlogit) package.

Many thanks for your help/thoughts!

randcoeff = list()
randcoeff = lapply(lX, function(x) eval(paste(text=paste0('mu_',x,'+sd_',x,'*draws_',x))))
names(randcoeff) = paste0('b_',lX)
return(randcoeff)
The error is about "e" needing to be a function/symbol/call/value
Top
stephanehess
Site Admin
Posts: 1042
Joined: 24 Apr 2020, 16:29

Re: Fully correlated MXL model

Post by stephanehess »

Hi

we are working on developing an interface, but it will take some time. Please see http://apollochoicemodelling.com/forum/ ... t=286#p958

Thanks
--------------------------------
Stephane Hess
www.stephanehess.me.uk
Top
cybey
Posts: 60
Joined: 26 Apr 2020, 19:38

Re: Fully correlated MXL model

Post by cybey »

Code: Select all

I've initially used the bayesian approach to "estimate" such model but there is something odd with the results (ie, the improvement in the loglik is way too large to be true) and when i use the bayesian approach to estimate an uncorrelated (diagonal) MXL model, the results don't match what i obtain with the "gmnl" (mlogit) package.
The LL reported in RSGHB is the LL at the posterior means, so using the lower level model. If you want to compare results with classical estimation, you should look at the upper level model. May this explain the differences in your results?

Regarding the differences between Apollo/RSGHB with Bayesian estimation and mlogit: Are you sure that your chains have converged, e.g. by using a sufficiently high number of draws? Differences can also occur if you used too few random draws in classical estimation, or your model simply converged to an inferior solution, which is more likely to happen the more complex your model is.

Nico
Top
Post Reply

Return to “Model specification”