Hey guys and girls,
I am new here. I am a new user of Stata, too. It is my second day. The reason for switching from SPSS to Stata is that I have to calculate an alternative-specific conditional logit model in my PhD thesis, which is not implemented in SPSS yet.
Trying to get familiar with the ASCLOGT procedures in Stata I played with the car choice example dataset. In reference to that i got some questions and I really hope someone you is able to help me:
The example model consists of case variables "sex" and "income" and the alternative-specific variable "dealer", see http://www.stata.com/manuals13/rasclogit.pdf.
If I have a look at the results as odd ratios Stata calculated alternative-specific odd ratios for all case-specific variables. So raising a person’s income increases the likelihood that the person purchases a Japanese or European car, for example. That is quite nice so far as well as clear.
In contrast to that I am not sure how to interpret the odd ratio for dealer. Should it tell me that raising the number of car dealers (no matter what type of car) increases the likelihood that a person purchases a car (once again: no matter what type of car), should not it?
Anyway, is there any possibility to get the odd ratios for alternative-specific variables alternative-specific like the case-specific variables, so that I could ascertain that a raising number of car dealers for European cars let increase the likelihood to purchase a European car maybe? I guess, it might be a quite interesting information as well, especially looking forward my PhD project.
Using the postestimation tools for ASCLOGIT leads to further knowledge, see http://www.stata.com/manuals13/rascl...estimation.pdf.
The estat mfx procedure calculates marginal effects of the model alternative-specific. Due to that I got few questions as well:
What kind of marginal effect is calculated there? Discrete changes? Marginal effects at the means? Average marginal effects? Or something else? But whatever there is calculated, how do I calculate discrete changes, marginal effects at the means and average marginal effects for a ASCLOGUT model in Stata? I did not find any working procedure.
In addition to that: I was not able to calculate Hausman's specification test or likelihood-ratio test, too. And I am missing McFaddens pseudo R2.
PS: What is the best way to avoid (and check) multicollinearity within an ASCLOGIT model using Stata?
Well, I know, lots of questions and some of them are stupid maybe, but I really hope someone is able to help me.
Thank you very much. I am looking to hear from you soon!
Best wishes from Marburg, Germany
Felix
I am new here. I am a new user of Stata, too. It is my second day. The reason for switching from SPSS to Stata is that I have to calculate an alternative-specific conditional logit model in my PhD thesis, which is not implemented in SPSS yet.
Trying to get familiar with the ASCLOGT procedures in Stata I played with the car choice example dataset. In reference to that i got some questions and I really hope someone you is able to help me:
The example model consists of case variables "sex" and "income" and the alternative-specific variable "dealer", see http://www.stata.com/manuals13/rasclogit.pdf.
If I have a look at the results as odd ratios Stata calculated alternative-specific odd ratios for all case-specific variables. So raising a person’s income increases the likelihood that the person purchases a Japanese or European car, for example. That is quite nice so far as well as clear.
In contrast to that I am not sure how to interpret the odd ratio for dealer. Should it tell me that raising the number of car dealers (no matter what type of car) increases the likelihood that a person purchases a car (once again: no matter what type of car), should not it?
Anyway, is there any possibility to get the odd ratios for alternative-specific variables alternative-specific like the case-specific variables, so that I could ascertain that a raising number of car dealers for European cars let increase the likelihood to purchase a European car maybe? I guess, it might be a quite interesting information as well, especially looking forward my PhD project.
Using the postestimation tools for ASCLOGIT leads to further knowledge, see http://www.stata.com/manuals13/rascl...estimation.pdf.
The estat mfx procedure calculates marginal effects of the model alternative-specific. Due to that I got few questions as well:
What kind of marginal effect is calculated there? Discrete changes? Marginal effects at the means? Average marginal effects? Or something else? But whatever there is calculated, how do I calculate discrete changes, marginal effects at the means and average marginal effects for a ASCLOGUT model in Stata? I did not find any working procedure.
In addition to that: I was not able to calculate Hausman's specification test or likelihood-ratio test, too. And I am missing McFaddens pseudo R2.
PS: What is the best way to avoid (and check) multicollinearity within an ASCLOGIT model using Stata?
Well, I know, lots of questions and some of them are stupid maybe, but I really hope someone is able to help me.
Thank you very much. I am looking to hear from you soon!
Best wishes from Marburg, Germany
Felix
Comment