Announcement

Collapse
No announcement yet.
X
  • Filter
  • Time
  • Show
Clear All
new posts

  • cmp command, When we use cmp for multivariate ordered probit model?

    Great day,
    When we use cmp for multivariate ordered probit model?

    My dissertation about adaptation strategies (5 strategies) for climate change by farmers.
    First step, I use Multivariate probit model, stata command:

    mvprobit (strategy 1 = independent variables) (strategy 2 = independent variables) (strategy 3 = independent variables)

    Strategy is dependent variable with value of 0 or 1. (0 is no use strategy; 1 is using strategy).
    Advantage: use mvprobit because we can analysis that farmers can use many adaptation strategies at the same time (simultaneously).
    Probit, Multinominal probit models did not consider the interrelationships among adaptation strategies at the same time.

    I want to ask everyone about Multivariate ordered probit model.
    Second step, I ask farmers about level for each strategy, odered for apply level , for example:
    strategy 1: (1) low level of apply, (2) medium level of apply, (3) high level of apply.
    strategy 2: (1) low level of apply, (2) medium level of apply, (3) high level of apply.
    strategy 3: (1) low level of apply, (2) medium level of apply, (3) high level of apply.

    is this multivariate ordered probit model? Can I use command?

    cmp (strategy 1 = independent variables) (strategy 2 = independent variables) (strategy 3 = independent variables), indicator (5 5 5)


    indicator (5 5 5) mean ordered probit for each equation.
    Advantage: use cmp, because we can consider strategies can use simultaneously in same time, strategies have interrelationships.
    If I use univariate probit, I have to consider each strategy, and I seem that strategies is independent.

    help me to use correct my above answer and how can I use cmp for using many strategies in same time (simultaneously). Because, use many strategies in same time for farmers is more reality.

    cmp, use for many equations occur in same time?

    Thank you


  • #2
    Chao Em Nam,
    Yes, that looks right to me. You can use cmp for both the multivariate probit and the multivariate ordered probit. For the multivariate probit model you could do:
    Code:
    cmp (strategy 1 = independent variables) (strategy 2 = independent variables) (strategy 3 = independent variables), indicator(4 4 4) nolr
    But because the mvprobit command is specialized for the multivariate probit, it may have more useful features for that model. For example, the post-estimation command mvppred generates predictions that are harder to generate with cmp (though still possible I think).
    --David

    Comment


    • #3
      Wow, Wow, Sir already taught for Fulbright program in Vietnam and worked for WB in Hanoi. That is why, Sir can understand my English, hahaha.
      Unfortunately, your article about cmp, published after you taught in Vietnam. I really like the teachers from US, because they are good at econometrics and many new econometric models.

      Sir can add priority to Vietnamese student. Sir help me to check and correct for 2 my answers:
      First answer:
      logit (probit), or oprobit (ologit) or mlogit (mprobit), these models will be applied when dependent variable is only 1 option for choosing (or many options but choose only 1 option), options must be independent (no correlation).

      With MVP or MOVP, we can select many options simultaneously, at the same time. Mean that, in MVP and MOVP, the options are correlation (i.e. dependent variables are correlate). Thus, before showing the results of model. We have to put correlation table for dependent variables. (pairwise correlation table to check correlation for dependent variables).

      Second answer:
      For MVP, we should use mvprobit, because it is specialized for MVP. After using mvprobit, we have command for joint probability (command like mvppred).

      For MOVP, we use cmp. But, how to calculate joint probability, conditional probability and marginal effect (unconditional probability)? When I use command: mvnormal , STATA said: command mvnormal is unrecognized, cry cry cry. Or Sir have article, write about how to calculate these probabilites.

      I think, this post about cmp, many students know more meaning about MVP and MOVP, then, they can more easily to apply for research, hihihihiihi.
      Thank you soooo muchhhhhhhhh Sir David.





      Comment


      • #4
        Hi Nam,
        Yes, you would use mlogit or mprobit (or cmp with the mprobit model) if only one alternative can be chosen/non-zero.

        Actually, for mprobit/mlogit the choices don't have to be independent, as long as there are "alternative-specific regressors" for the choices: each choice has some variable in its equation that not all of the others have. I talk about this in my paper on cmp. This is why Stata offers the asmprobit command. cmp can imitate asmprobit too: see the example in the help file.

        With some effort you can predict, and compute marginal effects on, a lot of different things after cmp, including joint probabilities of given outcomes. For more about this, look for the word "preserve" in the cmp help file.

        Good luck.

        --David

        Comment


        • #5
          Thank you very much Sir.
          your guidance help me to do more about cmp.

          Thank you Sir again.

          Comment

          Working...
          X