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.

Mean centering predictors on subject level

Ask general questions about model specification and estimation that are not Apollo specific but relevant to Apollo users.
Post Reply
apollo_user_1
Posts: 1
Joined: 24 Mar 2021, 10:11

Mean centering predictors on subject level

Post by apollo_user_1 »

Hi!

I have data from where participants could choose how to split 10 points between themselves and a random recipient. Before they had to do this, they had to rate how personally appropriate they think each action is, on a 4-point scale with numerical values -1, -1/3, 1/3, and 1. I use the payoff and the appropriateness rating as predictors in the conditional logit model.

What I discovered is that I get the same results when, instead of taking the raw appropriateness ratings, I take the deviation of a given actions appropriateness rating from the mean appropriateness rating of a given subject, i.e., personal norm rating of action i, subject j – mean personal norm rating over all actions of subject j.

I guess, but am not entirely sure, that this makes sense because I essentially just center the norm ratings on the subject level? And similar to centering a variable in a linear regression, the coefficient (excluding the intercept) does not change.

Following this train of thought, if the model fit gives a positive utility weight for the appropriateness rating, is it valid to say that on average (meaning average over all subjects), the subjects choose actions that have a higher appropriateness rating than the average appropriateness rating over all alternatives for a given subject (i.e., over all subjects, the average deviation (appropriateness rating of chosen action of subject j – mean appropriateness ratings of all actions of subject j) is positive)?

I made some plots that align with this interpretation:
apollo_forum.png
apollo_forum.png (14 KiB) Viewed 6402 times
The plots show the available alternatives on the x-axis and each point is a subject that ended up choosing that alternative. On the y-axis is how the appropriateness rating of the chosen alternative for a given subject j deviates from the mean appropriateness rating over all alternatives for that given subject j. The red line is the mean of all deviations over all subjects.

What I could show with the models is that if I exclude subjects who chose the alternative “exit”, the utility weight of the appropriateness rating is positive (significantly) (right plot), but when I include them it is a lot smaller (though still positive), but not significant anymore (left plot). The plots seem to confirm this (i.e., the mean deviation over all subjects (red line) is higher if we exclude subjects who chose the alternative “exit”).

Would appreciate it if you could tell me whether what I wrote makes any sense.
stephanehess
Site Admin
Posts: 974
Joined: 24 Apr 2020, 16:29

Re: Mean centering predictors on subject level

Post by stephanehess »

Hi

what you describe in the first part is just a reflection of only differences in utility mattering, so this is not surprising.

However, on a more general note, the use of this appropriateness rating as an explanatory variables raises important concerns about endogeneity bias. It would seem much more appropriate to treat this rating as a dependent variable alongside the allocation, in a hybrid choice model.

In relation to your final point, it's not quite clear what you mean by the exit option

Hope this helps

Stephane & David
--------------------------------
Stephane Hess
www.stephanehess.me.uk
Post Reply