David Heer, In Memoriam: Demography and Statistics

I remember arguing with David about the necessity of normalizing my variables by transforming them using standard procedures such as logarithmic transformation. True to his generation of applied statisticians, David did not see anything wrong with “normalizing” variables which were skewed, truncated, or otherwise not normally distributed. And I, also true to form, always the maverick, argued against it, feeling “instinctively” that information would be lost in the process of transforming raw variables which were distributed in their native form, that is, as originally observed.

Our intellectual dispute was rooted in the need to use the General Linear Model (GLM), specifically, multivariate numerical and categorical regressions to model and predict the impact of multiple spatial moves on cumulative fertility of Filipino women. The GLM, while possessing amazing versatility in modeling sociological phenomena like demographic processes, is hostage to the demands of parametric estimation: for its coefficients to be valid and reliable, the variable to be regressed and also its predictors must be normally distributed, and have (more or less) similar variances (homoscedastic).

Therefore, if not originally observed to be so, they must be transformed into more normal form by applying normalizing techniques. Some of the more common techniques are by transforming each value into its logarithm in base 10. In this way, an originally skewed distribution (which is not normal) will be linearized and normalized.

I remember once, while examining the frequency distributions of my variables, and running preliminary specifications, sitting alone in the Economics-Sociology computer laboratory, David unexpectedly walked into the room, and sat by me. He said he wanted to see my multivariate distributions, and perhaps my preliminary results. Now David works in the Population Research Laboratory as Deputy Director, but the laboratory was at least a mile away from the computer where I was working in Kaprielian Hall, in the main campus of the University of Southern California. Since David never drove a car to work but instead rode his racer bike twice daily, to and from work, he must have walked or biked from the Population Research Laboratory to get to me.

As conceived by Everett Lee, migration is a discrete variable: either a move occurs or it does not, and therefore a woman's record of moves necessarily skewed. There are no negative moves, and some women like those highly educated and employed as professionals move more than once. Like income, it has a skewed distribution. David therefore wanted me to normalize it before using it as a dependent variable in a multivariate model. Of course, current developments or more accurately, the current popularity of (already old) techniques like re-randomization, permutation and bootstrap as alternatives to logarithmic transformation were either unknown or unacceptable to David then.

Maybe later, I will write in detail about bootstrap or re-randomization as more powerful alternatives to logarithmic transformation, but now I want to acknowledge my debt and thus express my gratitude to a respected mentor, who took time from his busy schedule to attend to my academic needs. You see, David Heer, distinguished alumnus of Harvard University, author of numerous articles on demography and applied statistics, and of the world famous book, Society and Population, was my dissertation adviser.

Just a few months ago, I was thrilled at the prospect of conversing with him again after 17 years upon my return to the United States. I did not know at that time that I would forever be denied this pleasure. Unknown to me, and to me at least, for unknown reasons, David passed away several years ago (I do not even know when exactly). My attempts to locate information about his death from the Internet yield ed nil, and my attempts to contact his family through the phone failed. Understandably, in a long duration of 17 years, even area codes in Southern California could and did change. I have always been a natural researcher, and I could have turned earth over (only the surface, of course) to find my target information, but something told me to respect his family's (wish for) privacy. So I stopped trying to penetrate the social and cybernetic barriers about his death.

I do have some last words to say still, and for the first and last time, I am going to say them. I miss my student days at the USC, most of all my graduate work doing directed research (which makes up more than a third of my academic work at USC), all of which save the one in the Department of Geography under Curt Roseman were done under the supervision of David. Spanning the areas of migration, urbanization, economic development, environmental planning (with Tridib Banerjee), mortality (with Maurice Van Arsdol) and fertility, David always insisted on quantitative rigor, and the use of new statistical techniques. One set of techniques I particularly remember learning from him is event history analysis, and the Cox regression in particular. Never have I been so avid a positivist as I was under his supervision. Never had the use of the computer and statistical analysis been so delightful! [to my own students I say this: Weber and comparative historical sociology would thrill me much later]

I miss seeing David's unchanging light blue suit (he must have dozens of them) coming into the Population Research Laboratory bringing in his bike at the beginning of the work day. After receiving my own doctorate, I remember him writing an invitation to present my findings on Filipina migrants to Germany to a USC audience, which did not materialize. That trip to Germany quickened a latent love for Weberianism, and my interest in the developmental state began to emerge as a dominant theme in my teaching and research. I noticed in myself a pull toward German political sociology, and Germany as a model system for a politics and culture.

In the German academic tradition, the dissertation adviser is akin to a father, who raises a son academically. To ease my regret at not having expressed in person my gratitude for what he has done for me, I say, Sir Professor Doktor David Heer, mein DoktorVater, Ich liebe Dich.