the dependent variables are adopters of synthetic fertilizers , non-adopters of synthetic fertilizers and those households that use both cattle manure and synthetic fertilizers
Dear Laston,
A simple model can treat these three variables as inputs, as three method to produce something. For example, if you take a timeseries approach, then you can have dummies for each variable to explain some yield variable. For instance
Yield = f(dummies for each method,other variables).
Take cares of three things:
1. Multicolinearity
2. Dummy variable trap
3. Keep the observation in time order to be able to detect serial or auto correlation.
Good luck in your research.
respected Lall B. Ramrattan sir
is it possible to take more than one dependent variable in panel data analysis ? suppose if i have three dependent variable and one independent variable which model suites well ?
Dear Darshini S
Please see this link,
http://www.psych.yorku.ca/lab/psy6140/lectures/MultivariateRegression2x2.pdf
I think that, rather than three dependent variables, you have one qualitative dependent variable that can take three diferent values (users of synthetic fertilizers; non-users of synthetic fertilizers and users of synthetic fertilizers that also use cattle manure). If this is the case, you should use a model for qualitative dependent variables. In this case, I would use a multinomial logit model. Most textbooks explain about this model (Greene has a very comprehensive explanation, but it might be too dense for those who are not used to Econometrics), and it is implemented in most statistical and econometric packages (SPSS, SAS, Limdep, or Gretl if you prefer "free" software).
Just to add, it all depends on the purpose of your model and the relationship you want to investigate, and sometimes data availability could be one issues as well. Generally, as explained by Lall B. Ramrattan, thats the basic simple model to start with but again if you have series or a panel (See woodridge or Gujurat on this). Then you might need to fit a time series or panel data model.
One suggestion can be a multinomial logit 0=non synthetic 1= synthetic and 2 =both. You can choose a reference category (any of them) to compare results. In this case you'll need some covariates /factors. Perhaps you want to measure the impact of the fertilizators on the production. Then you may use the three mentioned variables as determinants (x) -each as binary. If the latest is the goal keep in mind what Lall B. Ramrattan sayed. You can solve this problem using any econometric software you want. References mentioned above by other colleagues are very useful (Gujarati eg-Basic Econometrics or Econometrics by example )Good luck!
@ Darshini. You problem seems like a simultaneous equation problem.
@ Laston: Use, an OLS model and check for its assumptions of autocorrelation, multicollinearity, heterosckedasticity, etc.
The study is using cross-sectional data.The objective for the econometric model i asked is ;
To determine the factors influencing the use of synthetic fertilizers by smallholder farmers. I used the multinomial logit model,the issue now is on interpretation of the results of the multinomial model.Do I use the co-efficients,p- values or the marginal effects?
The coefficients of a Logit model are not directly interpretable. Most econometric programs show, for each coefficient, the marginal effect for an average individual, which is easy to understand and interpret. You can also use the odd-ratio for each variable (e exponetiated to the variable's coefficient), which tells you the marginal change in the odds [p/(1-p)], which is constant.