actual vs predicted value

[EL KIMON]

Dear all,
i want to compare the actual with the predicted mean hours. I run the following model:

mixlogit didep hours if sex==0 & marital_status==1, rand(disposable_income) id (idperson
> ) group(strata)

Iteration 0:   log likelihood = -786.53313  (not concave)
Iteration 1:   log likelihood = -595.57084  (not concave)
Iteration 2:   log likelihood = -368.88205  (not concave)
Iteration 3:   log likelihood = -326.03393  
Iteration 4:   log likelihood = -324.29482  
Iteration 5:   log likelihood = -322.96199  
Iteration 6:   log likelihood = -322.80673  
Iteration 7:   log likelihood = -322.80599  
Iteration 8:   log likelihood = -322.80599  

Mixed logit model                                 Number of obs   =       1455
                                                  LR chi2(1)      =       0.00
Log likelihood = -322.80599                       Prob > chi2     =     0.9875

------------------------------------------------------------------------------
       didep |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
Mean         |
       hours |  -.1618746    .010528   -15.38   0.000    -.1825091   -.1412401
disposable~e |   .0035452   .0003435    10.32   0.000     .0028719    .0042186
-------------+----------------------------------------------------------------
SD           |
disposable~e |   7.67e-06   .0004909     0.02   0.988    -.0009546    .0009699
------------------------------------------------------------------------------ 

If i use the mixlpred.. i get the didep_hat .
How i get the predicted mean hours?

Thank you.

[Maarten Buis]

In your model you use hours to predict didep, so hours is treated as given and thus cannot be predicted from your model.

[EL KIMON]

The hours is a variable that takes the values 10, 20, 30 ( 3 choices) and the variable didep is a categorical variable that takes the value 1 for the alternative of the 3 choices of the hours.
So by the model i get the prediction for the didep?
Is there any way to get the prediction of the hours?

[Maarten Buis]

With models like the one you are using there is typically a dependent/explained/left-hand-side/y-variable and one or more independent/explanatory/right-hand-side/x-variables. In your model hours is an explanatory variable, so it is considered as given. If something is given it cannot be predicted other than trivially use an exact copy of the original variable. So the answer to your questions are yes, it predict didep, and no hours cannot be (meaningfully) predicted.

[EL KIMON]

I didn't explain my model correctly i think.
The model that i wanted to estimate is the discrete labour supply model following van Soest (1995), Aaberge et al.(1995, 1999), Hoynes (1996), Blundell et al. (2000), Eissa and Hoynes (2004) Eissa et al. (2008).
we adopt the flexible quadratic utility function U(H,Y, Z) where h are the hours and y the income.
As we estimate a discrete choice model, we must first decide the finite set of hours according to which individuals choose their hours. For each choice hi we have the analogous disposable income yi.

To estimate the model we must add stochastic terms to the utility function. In what follows, we only add shocks specific to the state or hours regime for each of the possible choices, which we assume are generated by extreme value distributions. Following these assumptions, we derive the choice probability for agent i.
I used the mixelogit to estimate all the parameters in the model.

[Maarten Buis]

The code you gave us and the description you just sent us are so far removed from one another that I don't think it is possible to help you through a forum like this. Forums like these are great, but like any form of comunication they have their weak points; Forums are good at small quick questions, but they don't work well when extensive consulting/teaching is needed. What you need is face to face consulting. So I suggest you find someone local who can help you.