Hi, I have a DCE dataset, where respondents chose eight times between two options. Each option had six attributes with 3 attribute levels. Five of these 6 are effects-coded and the last one is dummy-coded. I ran lclogit with 4 classes, which gives me the following output:
I then used lclogitml, which initialises gllamm and gives the following results for class 1:
Now, what I want is simply to get marginal effects / predicted probabilities for the coefficients (which I would get in simpler models such as logit via "margins, dydx(*)" ).
I passed the data to gllamm via "lclogitml, switch", and tried various things with "gllapred" and "gllasim". However, I can't figure out this rather simple thing.
Code:
lclogit choice ch1_eff1 ch1_eff2 ch2_eff1 ch2_eff2 ch3_eff1 ch3_eff2 ch4b_eff1 ch4b_eff2 ch5c_eff1 ch5c_eff2 ch6_priv_dummy ch6_pub_dummy, id(id) group(cc2) nclasses(4)
Code:
-------------------------------------------------
Variable | Class1 Class2 Class3 Class4
-------------+-----------------------------------
ch1_eff1 | -1.219 -0.828 -0.536 -0.004
ch1_eff2 | 0.183 1.028 0.065 -0.092
ch2_eff1 | -2.243 -0.673 -2.235 0.072
ch2_eff2 | 2.131 0.310 1.953 -0.204
ch3_eff1 | -2.576 -1.006 -0.959 0.285
ch3_eff2 | 2.840 0.635 0.471 0.330
ch4b_eff1 | -0.229 -0.466 -0.653 -0.016
ch4b_eff2 | 1.043 0.237 0.500 -0.154
ch5c_eff1 | -1.946 0.236 -0.548 -0.918
ch5c_eff2 | 1.419 -0.135 0.534 -0.163
ch6_priv_d~y | -1.562 -0.044 -0.087 -0.410
ch6_pub_du~y | 0.507 -1.350 -0.251 0.813
-------------+-----------------------------------
Class Share | 0.325 0.380 0.179 0.115
-------------------------------------------------
Code:
--------------------------------------------------------------------------------
choice | Coefficient Std. err. z P>|z| [95% conf. interval]
---------------+----------------------------------------------------------------
choice1 |
ch1_eff1 | -1.21861 .2341479 -5.20 0.000 -1.677531 -.7596882
ch1_eff2 | .1832757 .0839117 2.18 0.029 .0188117 .3477396
ch2_eff1 | -2.242841 .3223866 -6.96 0.000 -2.874707 -1.610975
ch2_eff2 | 2.13119 .3397402 6.27 0.000 1.465311 2.797068
ch3_eff1 | -2.575844 .2938692 -8.77 0.000 -3.151817 -1.999871
ch3_eff2 | 2.839723 .3385522 8.39 0.000 2.176173 3.503273
ch4b_eff1 | -.2288113 .1535203 -1.49 0.136 -.5297055 .072083
ch4b_eff2 | 1.043092 .2627628 3.97 0.000 .5280866 1.558098
ch5c_eff1 | -1.945933 .3752163 -5.19 0.000 -2.681344 -1.210523
ch5c_eff2 | 1.418596 .3618886 3.92 0.000 .7093076 2.127885
ch6_priv_dummy | -1.562217 .212733 -7.34 0.000 -1.979166 -1.145268
ch6_pub_dummy | .5069582 .1806134 2.81 0.005 .1529624 .860954
---------------+----------------------------------------------------------------
Now, what I want is simply to get marginal effects / predicted probabilities for the coefficients (which I would get in simpler models such as logit via "margins, dydx(*)" ).
I passed the data to gllamm via "lclogitml, switch", and tried various things with "gllapred" and "gllasim". However, I can't figure out this rather simple thing.
