Ranked ordered model

[Jean Dewals]

Dear collegues,

I'm currently working on a database from a consumers survey. In this survey, we asked individuals to rank their top three preferred labelling schemes out of nine alternatives.

Now I would like to set up a ranked ordered model that allows me to explain the preferences of our individuals with regard to several variables (gender, age, income ....)

My database is organised as follow :

* Example generated by -dataex-. For more info, type help dataex
clear

input byte ID str6 Label_name byte(pref male deplittoral) double univ
 1 "local"  3 0 0  1.304165
 1 "france" 0 0 0  1.304165
 1 "equi"   0 0 0  1.304165
 1 "euro"   0 0 0  1.304165
 1 "envi"   0 0 0  1.304165
 1 "qual"   2 0 0  1.304165
 1 "sante"  1 0 0  1.304165
 1 "nutri"  0 0 0  1.304165
 1 "anx"    0 0 0  1.304165
 2 "local"  3 0 0  1.029258
 2 "france" 0 0 0  1.029258
 2 "equi"   0 0 0  1.029258
 2 "euro"   0 0 0  1.029258
 2 "envi"   2 0 0  1.029258
 2 "qual"   0 0 0  1.029258
 2 "sante"  1 0 0  1.029258
 2 "nutri"  0 0 0  1.029258
 2 "anx"    0 0 0  1.029258
 3 "local"  2 0 1 -1.549724
 3 "france" 1 0 1 -1.549724
 3 "equi"   0 0 1 -1.549724
 3 "euro"   0 0 1 -1.549724
 3 "envi"   0 0 1 -1.549724
 3 "qual"   3 0 1 -1.549724
 3 "sante"  0 0 1 -1.549724
 3 "nutri"  0 0 1 -1.549724
 3 "anx"    0 0 1 -1.549724
 4 "local"  0 0 0  .9819833
 4 "france" 2 0 0  .9819833
 4 "equi"   0 0 0  .9819833
 4 "euro"   0 0 0  .9819833
 4 "envi"   3 0 0  .9819833
 4 "qual"   0 0 0  .9819833
 4 "sante"  0 0 0  .9819833
 4 "nutri"  0 0 0  .9819833
 4 "anx"    1 0 0  .9819833
 5 "local"  0 0 1  .7331646
 5 "france" 0 0 1  .7331646
 5 "equi"   0 0 1  .7331646
 5 "euro"   0 0 1  .7331646
 5 "envi"   0 0 1  .7331646
 5 "qual"   0 0 1  .7331646
 5 "sante"  3 0 1  .7331646
 5 "nutri"  2 0 1  .7331646
 5 "anx"    1 0 1  .7331646
 6 "local"  0 1 0  .7931689
 6 "france" 0 1 0  .7931689
 6 "equi"   3 1 0  .7931689
 6 "euro"   0 1 0  .7931689
 6 "envi"   1 1 0  .7931689
 6 "qual"   0 1 0  .7931689
 6 "sante"  0 1 0  .7931689
 6 "nutri"  0 1 0  .7931689
 6 "anx"    2 1 0  .7931689
 7 "local"  3 0 1  .7203799
 7 "france" 0 0 1  .7203799
 7 "equi"   2 0 1  .7203799
 7 "euro"   0 0 1  .7203799
 7 "envi"   0 0 1  .7203799
 7 "qual"   0 0 1  .7203799
 7 "sante"  0 0 1  .7203799
 7 "nutri"  0 0 1  .7203799
 7 "anx"    1 0 1  .7203799
 8 "local"  0 0 0 -.5312917
 8 "france" 1 0 0 -.5312917
 8 "equi"   0 0 0 -.5312917
 8 "euro"   0 0 0 -.5312917
 8 "envi"   2 0 0 -.5312917
 8 "qual"   0 0 0 -.5312917
 8 "sante"  0 0 0 -.5312917
 8 "nutri"  0 0 0 -.5312917
 8 "anx"    3 0 0 -.5312917
 9 "local"  0 1 1 -.3313667
 9 "france" 0 1 1 -.3313667
 9 "equi"   0 1 1 -.3313667
 9 "euro"   2 1 1 -.3313667
 9 "envi"   3 1 1 -.3313667
 9 "qual"   1 1 1 -.3313667
 9 "sante"  0 1 1 -.3313667
 9 "nutri"  0 1 1 -.3313667
 9 "anx"    0 1 1 -.3313667
10 "local"  0 0 0  1.206946
10 "france" 0 0 0  1.206946
10 "equi"   0 0 0  1.206946
10 "euro"   0 0 0  1.206946
10 "envi"   2 0 0  1.206946
10 "qual"   0 0 0  1.206946
10 "sante"  3 0 0  1.206946
10 "nutri"  0 0 0  1.206946
10 "anx"    1 0 0  1.206946
11 "local"  0 1 0 -.3468108
11 "france" 0 1 0 -.3468108
11 "equi"   0 1 0 -.3468108
11 "euro"   0 1 0 -.3468108
11 "envi"   1 1 0 -.3468108
11 "qual"   0 1 0 -.3468108
11 "sante"  2 1 0 -.3468108
11 "nutri"  0 1 0 -.3468108
11 "anx"    3 1 0 -.3468108

end

Some explanations :

• The first column "ID" represents each indivdiuals (here, we have the ranking of 11 individuals)
• The second column "Label_name" represents the 9 possible choices of labelling schemes
• The third column "pref" represents the ranking given by the individual for the alternative considered in the previous column. The 1 represents the preferred alternative. As we asked to rank the 3 preferred alternatives for each individual, the 0s represent the unranked alternatives. Each individual therefore has 6 unranked alternatives, so six zeros.
• The fourth column indicates the gender of the individual (1 = Female ; 0 = male)
• The fifth column indicates if the individual lives in a coastal departement
• The last colum represents the degree of altruism of the individual

I chose to use a cmroprobit model. regarding my data and my purpose. I organised my code as follows

cmset ID Label_name
cmroprobit pref, casevars (male deplittoral univ)

I wanted to know if you think the model I use is the most relevant? Or would another model allow me to obtain more interesting results?

I hope that my explanations are clear enough, I remain available to bring you additional information if necessary !

Best regards,

DEWALS Jean-François