如何在混合 logit 模型中指定属性级别？

问：

我正在进行一项关于消费者偏好的研究，要求受访者根据一组属性选择产品。在运行混合 logit 模型后，我得到了以下结果。我正在寻求有关如何按属性级别隔离结果的帮助。

. mixlogit y price3, rand(texture ingrid) group(option_set) id(numeric_id) nrep(200) 

Iteration 0:   log likelihood = -689.71115  (not concave)
Iteration 1:   log likelihood = -685.58262  
Iteration 2:   log likelihood = -683.87879  
Iteration 3:   log likelihood = -683.84319  
Iteration 4:   log likelihood = -683.84307  
Iteration 5:   log likelihood = -683.84307  

Mixed logit model                                       Number of obs =  3,096
                                                    LR chi2(2)    =  11.55
Log likelihood = -683.84307                             Prob > chi2   = 0.0031

------------------------------------------------------------------------------
       y | Coefficient  Std. err.      z    P>|z|     [95% conf. interval]
-------------+----------------------------------------------------------------
Mean         |
  price3 |  -.0009266   .0001381    -6.71   0.000    -.0011973   -.0006558
 texture |  -1.174523   .0879522   -13.35   0.000    -1.346906    -1.00214
  ingrid |  -.1090069   .0652664    -1.67   0.095    -.2369268    .0189129
-------------+----------------------------------------------------------------
SD           |
 texture |   .6666297    .130163     5.12   0.000      .411515    .9217444
  ingrid |   .0030574   .1314769     0.02   0.981    -.2546326    .2607474
------------------------------------------------------------------------------`

我尝试使用运算符，它通常用于多项式对数分析。但是，我遇到了一个错误，指示我正在执行不允许的时间序列操作。i.

. mixlogit y price3, rand(i.texture i.ingrid) group(option_set) 
id(numeric_id) 
nrep(200) 
factor-variable and time-series operators not allowed
(error in option rand())

我尝试将属性转换为虚拟变量，但由于共线性，模型最终删除了大部分变量。

我希望结果类似于下面介绍的结果。不应使用条件 logit 模型，而应采用混合 logit 模型。

. clogit y price3 i.texture5 i.ingrid5, group(numeric_id) 
note: 5.ingrid5 omitted because of collinearity.
note: multiple positive outcomes within groups encountered.

Iteration 0:   log likelihood = -2132.9689  
Iteration 1:   log likelihood = -2132.5788  
Iteration 2:   log likelihood = -2132.5786  

Conditional (fixed-effects) logistic regression        Number of obs =   9,288
                                                       LR chi2(6)    = 1402.42
                                                       Prob > chi2   =  0.0000
Log likelihood = -2132.5786                            Pseudo R2     =  0.2474

------------------------------------------------------------------------------
           y | Coefficient  Std. err.      z    P>|z|     [95% conf. interval]
-------------+----------------------------------------------------------------
      price3 |  -.0016488   .0001815    -9.09   0.000    -.0020044   -.0012931
             |
    texture5 |
     coarse  |   1.692396   .6769401     2.50   0.012     .3656177    3.019174
       fine  |   2.705765   .6652764     4.07   0.000     1.401847    4.009683
             |
     ingrid5 |
 70,20,0,10  |   .9481876   .1057004     8.97   0.000     .7410186    1.155357
  100,0,0,0  |  -1.167767   .1452245    -8.04   0.000    -1.452401 -.8831319
 70,15,10,5  |   .5271418   .1093191     4.82   0.000     .3128803    .7414032
  Other  |          0  (omitted)
------------------------------------------------------------------------------