Hello,
I'm currently trying to examine the effects of export promotion on export market survival.
In doing so, I am going to use complementary log-log model (cloglog).
My first question is as follows:
I understand that cloglog model is appropriate in case of "discrete" time (i.e. year).
I do have yearly data (regarding export status of firms, export promotion, etc.), but unfortunately, there are some gaps in the years.
For example, I only have data for 2008, 2009, 2011, and 2013. I was thinking that cloglog model would not work because the data are not truly yearly.
In this situation, is it still possible for me to use cloglog model?
Alternatively, would it be okay if I use biennial data (2009, 2011, and 2013)?
My second question is regarding a command.
In case of nonlinear models including probit or logit, using commands such as "mfx" or "margins" was extremely convenient. Are there any commands which compute the probability of exit from export markets evaluated at the mean (of an independent variable)?
My third question may be related to the second question. The independent variables that I consider using include dummies, ratios (ranging from 0 to 1), and natural logarithms. I am curious of how to interpret coefficients on such variables.
In case of a coefficient b on on an independent variable X, I think that a unit increase in X would increase the probability of exit from export markets by {exp(b)-1} percentage points.
However, I suspect that interpretation might differ whether X is a dummy, a ratio, or a natural logarithm.
My last question is regarding "multiple spells."
Obviously, in the data, some of the firms enter export markets for the first time, exit, and then re-enter the export markets, which leads to multiple spells. Is there any way that I can solve this issue when I use the cloglog model?
Thank you very much for your time to read my questions, and I would very much appreciate it if anyone could help me out on these issues.
I'm currently trying to examine the effects of export promotion on export market survival.
In doing so, I am going to use complementary log-log model (cloglog).
My first question is as follows:
I understand that cloglog model is appropriate in case of "discrete" time (i.e. year).
I do have yearly data (regarding export status of firms, export promotion, etc.), but unfortunately, there are some gaps in the years.
For example, I only have data for 2008, 2009, 2011, and 2013. I was thinking that cloglog model would not work because the data are not truly yearly.
In this situation, is it still possible for me to use cloglog model?
Alternatively, would it be okay if I use biennial data (2009, 2011, and 2013)?
My second question is regarding a command.
In case of nonlinear models including probit or logit, using commands such as "mfx" or "margins" was extremely convenient. Are there any commands which compute the probability of exit from export markets evaluated at the mean (of an independent variable)?
My third question may be related to the second question. The independent variables that I consider using include dummies, ratios (ranging from 0 to 1), and natural logarithms. I am curious of how to interpret coefficients on such variables.
In case of a coefficient b on on an independent variable X, I think that a unit increase in X would increase the probability of exit from export markets by {exp(b)-1} percentage points.
However, I suspect that interpretation might differ whether X is a dummy, a ratio, or a natural logarithm.
My last question is regarding "multiple spells."
Obviously, in the data, some of the firms enter export markets for the first time, exit, and then re-enter the export markets, which leads to multiple spells. Is there any way that I can solve this issue when I use the cloglog model?
Thank you very much for your time to read my questions, and I would very much appreciate it if anyone could help me out on these issues.
Comment