Hello everyone !

I am currently working on a study of French consumers' preferences towards different types of seafood labels. In this study, we asked about 1200 individuals to rank their 3 favourite labels among a choice of 9 alternatives.

I first started working on logit and Clogit models in order to study the preferences of my consumers for each of the labels, without taking into account the assigned ranking. To this end, I created dichotomous (0/1) variables for each of the labels tested. 1 = label ranked by the individual, 0 = label not ranked by the individual. These models are have already been implented and analysed.


My goal is now to integrate the ranking of the labels into my analysis. I have therefore reworked my database in the following way (for the first 4 individuals).
Array


Some explanations about the variable in this screen:
  • ID corresponds to an individual
  • Label corresponds to the different labels tested in our survey
  • rank_lbl corresponds to the ranks attributed by the individual
  • gender corresponds to sex
  • recode~e to age categories
  • prefgeo to indivuals preferences for local products
  • Etc ...


In my opinion, my dataset is quite similar to panel data, with a very limited time step between the individual's answers.

To perform my analysis, I decided to work on a ranked ordered panel models, and I was interested in the xtologit model. This is the first time I have worked on this type of model, and I have several questions about it.
  1. Firstly, does my data fit well with the choosen model ? Would another model be more relevant according to you regarding my objectives ?
  2. Secondly, I wanted to know if this model was adapted to a draw without replacement? Here, when the consumer chooses his label in position 1 (his favourite), he can no longer rank it second (in rank 2). In other words, he chooses a first time among 9 choices, then among 8 choices, then among 7 choices. All remaining labels are then considered as not chosen. Is this a problem for the model I want to test?


I am also really interested in any advice on this type of model, any aditionnal informations on this model !

If my explanations are not clear, I remain available to provide you with any additional information

Thank you for your time,

Best regards,

Dewals Jean-François, PhD Student working on consumers preferences