Dear Friends,
I am currently working on this project. This has to be completed within two years and fully funded by the government of India body (ICSSR). Please read this proposal. You all are kindly requested to provide me the insights on the following:
1. previous Research Papers on the construction of the whistle blowing Index
2. seminal studies on whistleblowing
3. Proposed methodology is appropriate or not?
4. Proposed regression model (Logit) is the best fitted or not?
any other related suggestions are welcome.
thanks in advance.
Dr. Shweta
Dear Shweta,
Logistic regression models are usually used for dichotomous dependent variables(present/absent, yes/no, success/failure etc) with single or more independent variables. If you are categorizing your whistle blowing index into just two categories then you are good to go! but if you want more than two categories then you need another methodology to reach your objectives.
Hi! Thanks for your kind repl. most of the independent variables are dichotomous but few variables including control variables like firm size and leverage could be of different kind. In such situatio will that model be suitable?
Thanks.
In your proposal, you stated that you would be employing Tobit Regression (TR), but in your question, you wonder whether Logistic Regression (LR) can be used. Are you planning on building a model with a continuous dependent variable or a binary one? In the former case, you can proceed with TR or even Multiple Linear Regression (MLR) - based on your interest in including data with incomplete response. If your dependent variable (whistleblowing policy disclosure score) is dichotomized, then you can use LR; if it has more than two categories, Multinomial or Ordered LR models are available among several others. Note that in all these cases, you will be estimating pooled models, but I understand that you have a panel data: 10 years and 200 firms. If you want to measure the cross-sectional and longitudinal effects with a limited dependent variable, you should take a look at "panel data models for binary choice." By the way, what type of independent variables you include in logit-like models does not matter. You can even consider interaction effects. The coefficient estimates are interpreted differently compared to MLR though. Best of luck!
- Badges
- Science topic
- Similar topics
- Methodology
- Research Methodology
- Research Papers