So some of the variables were binary, some ordinal, and some continuous. Then we did exploratory principal components analysis, but didn't try to model each statutory choice with additional independent variables, as I think IRT does (my familiarity with IRT is hazy). Because of the large number of variables (over 170), the eigenvalues were grossly inflated, so I used Horn's Parallel Analysis to correct for random correlation among the variables, and once I did so, found that there were 2 eigenvalues well over 1.0 (the top adjusted eigenvalue over 10), and 1 eigenvalue just above 1.0. A scree plot also showed essentially 2 factors. The 3rd factor I can't make theoretical sense of (seems to capture some similarities between Oregon and Alaska), and its Horn-adjusted eigenvalue is marginal, so I think it's probably just a fluke. The data I present here are for the unrotated factors, which actually seem to make the most theoretical sense.
[warning - geeky response to follow]
Jason,
It appears your task is largely exploratory, and I can see the benefit of the exploratory factor analysis to that task. Rasch (same mathematically as 1PL IRT, but a different paradigm) has excellent techniques for dealing with mixed types of items (e.g. binary, rating scale, partial credit), but it requires you to have an apriori construct with items you hypothesize at each "inch" of the full construct of interest. You might consider a followup study to build out the scale more effectively using this modern psychometric approach. Some big advantages to Rasch/IRT:
a) You get interval-scale data, and if you can create an absolute zero, you get Ratio-scaled data (e.g. like physicial science)
b) Smaller, more consistent standard errors across the full range of the dimension you're measuring
c) Where applicable, you can use computer-adaptive methods to shorten questionnaires, surveys, tests, etc by 20-50%, once you know where each question falls on the liberty "ruler" of interest
on a different, but related thought -
Jason, as you're a proper academic, and I'm an Industrial/Organizational Psychologist (Chief Learning Officer/VP at a Healthcare system)....It would be fun to discuss a virtual "Free State University" that could have both You Tube-style liberty courses; and "Web-ex" style symposia, colloquia, and lectures. Perhaps it could be degree granting. I wonder with many FSP'ers looking to NH for a better freedom culture, if there is a growing freedom education marketplace for undergrad or graduate education for private and home schoolers.
Reactions anyone?
Matt