I am writing a proposal for my new research project and wondering whether any new powerful techniques have been recently introduced to measure the connectedness between factors and decision bilaterial variable (0/1)?
Perhaps it would help to add some clarification. When I read your question literally, it sounds like you are planning to work with a limited dependent variable using something like Probit analysis. For example, creating a model that might estimate probability of a recession in the US might have a recession as
1 = time period when a recession ocured
0 = otherwise.
Once again, your question reads like you anticipate a limited dependent variable. Instead of estimating a limited dependent variable, my guess is that you were more focused on using one or more limited independent variables (e.g., a dummy variable) along with other macroeconomic variables to estimate a dependent variable. This might be something like estimating a nation's labor force participation rate (or any other macroeconomic-type variable). In this example, a researcher might wish to use a dummy variable like this:
1 = a time period when the nation is involved in a war
0 = otherwise.
Obviously, war years may create a situation where controlling for it (i.e., war) might improve the model.
I suspect you will get more helpful answers if you clarify what you are thinking. Moreover, it probably would help to share a few more details about what you are thinking of exploring. I hope this helps and encourages others to answer.
you could run the logit/probit regression for your study. basically depending on the characteristics of the dependent variable your model might change. I would suggest you to read the limited dependent modelling techniques for your study.
I might be reading between the lines on Rob Catlett's comment, but it seems like it might be possible to create a quantitative variable even when you think that you only can use a qualitative variable. For example, what is the extent of the recession? Or what is the severity of the war?
Peter, you are correct, of course and in general. but not always, a continuous variable of at least a more robust one often provides valuable information that may extend beyond the limited 0 or 1 options. I was curious why Larisa seemed more interested in a binary variable and did not ask. So, you were reading between my lines correctly. Moreover, it sounded to me like she might be thinking of using it as the dependent variable. Thanks for clarifying. Rob
Thank you very much for your comments. I am actually using 0/1 variable as a dependent variable, and economic factors as regressors. Briefly, the research question is whether the changes in economic fundamentals affect on multinational corporations' decision making process, i.e. decision to stay in the country (0), or decision to leave (1) from this country.
We employed ReLogit regression analysis already, but I think that further tests are required to increase the reliability of the results. I will read about limited depending modelling.
Larisa, forgive me please for being a nag. It seems to me that a multinational almost never totally leaves. Instead, it relocates some portion of its activities. That is, this could be a quantitative variable . . . unless you are only interested in the headquarters of the multinational. Still, I would argue that the distinction between headquarters functions and back-office or regional functions may be arbitrary.
We are using data for subsidaries in stressful North African region, and both qualitative and quantitative variables are used. There are evidence of multiple exit decisions due to political violence in the region. (List of regressors is not-limited to macroeconomic factors though). This question is related to one very specific part of the data analysis, i.e. the determinants of stay-or-exit decision. Nevertheless, there is no issue with a story itself, the econometrics employed is what concerns me the most in this project. Therefore I asked for some methodological suggestions. So far ReLogit procedure has been used.