Dear Gary,
I hate to bother you with some questions, but I am having no luck getting an answer from anyone else.
Friends are sending me articles that are supposed to answer my questions, but they do not. I recently attended a talk you gave
At my school-Univ. of South Carolina. As with every other talk I have heard from you, I was motivated to make a change and do things differently. Unfortunately, I am stuck with some questions that no one seems able to answer.
I do large N analysis of wars, where most observations are zeroes. Your recent talk at USC was about matching and it sounded like RELogit had multiple benefits, including creating a more balanced sample. My understanding was the sample analyzed would consist of an equal number of events and non-events. . I hope I have interpreted this correctly.
That is not the main reason for my email to you. I have been struggling all week to try to figure out whether I really need to use Fixed or random effects in my analysis. As you know, I cannot run them in the RELogit program, but it seems like it is something reviewers want to see. I wondered whether there is a reason you excluded them from your program-for example, because you thought they were unnecessary with your program. I know that it is still possible to use clustering, but that does not do the exact same thing as fe and re.
I would appreciate any guidance about your thinking about my using RELogit for my main analysis and not using fe or re.
Thank you in advance for your assistance. I really appreciate it.
Best wishes,
Katherine Barbieri
Katherine Barbieri
Associate Professor
Department of Political Science
University of South Carolina
Columbia, SC 29205
Email: Katherine.barbieri(a)sc.edu<mailto:Katherine.barbieri@sc.edu>
Tel: (803) 467-3912 (cell); (803) 777-6902 (office)
http://people.cas.sc.edu/barbierk/