Dr. Alexander W. Schmidt-Catran

Catsem - A Stata ado for categorical data analysis

Together with Hans-Jürgen Andress and Maximilian Hörl. We are working on a stata ado that allows estimating a variety of statistical models by means of path notation. On the one hand, our command attempts to bring together a variety of routines that already exists scattered over different Stata commands. On the other hand, our ado adds new functionality to Stata: latent class analysis. The project started in December, 2014.

Coding Schemes for Categorical Variables: Weighted Effect Coding

Together with Manfred te Grotenhuis, Ben Pelzer, Rob Eisinga and Rense Nieuwenhuis. In this project we write a paper on weighted effect coding in regression analysis with categorical predictor variables. While effect coding is a well-known standard parameterization in regression analysis, weighted effect coding is much less prominent than it deserves. We furthermore propose a new way of coding interaction effects within the weighted effect coding framework. "Our" interaction effects are orthogonal to the main effects in the model. The paper is accompanied by a set of macros which implement the described techniques in various statistical programmes. I write an ado for Stata. The project started in February, 2015.

(Non-)Invariance in Welfare Attitude Measurements?

Together with Nate Breznau. In this paper we discuss the problem of measurment non-invariance in the welfare attitude module of the European Social Survey Round 4. We propose a multilevel measurement model that accounts for a country-level latent variable. When this variable is controlled for, measurment invariance holds for the measurement model.

Handbook of Comparative Social Research

Together with Hans-Jürgen Andress, we plan to edit a book on comparative research in social science. The book covers a variety of subjects, starting from research designs for comparative studies, over measurements issues in cross-national settings, to methods for the estimation of contextual-level effects. The project was initiated in December 2013. In the first phase of the project, we will determine the content of the book and propose an overall framework for the single chapters. In the next phase we will approach potential authors from the international scientific community and invite them to contribute to the book. I will contribute two chapters myself. One chapter will deal with advanced multilevel modeling techniques and a second chapter will address issues of robustness tests for multilevel models.

The Reciprocal Relationship between Social Policy and Public Opinion towards Welfare

There is a huge amount of literature that investigates the relationship between public opinion and public policy. Two approaches to this relationship are dominant in the scientific discussion. Many scholars assume that public opinion is shaped by the political and institutional context. The basic assumption of this research is that the institutional setting, in which individuals form their opinions, has a normative effect on the dominant logic of solidarity. The theoretical idea that policies influence peoples' opinion is often termed policy feedback. Other scholars belief that public opinion influences policy makers as one should expect in a representative democracy. This theoretical approach is often called opinion representation. The major shortcoming of previous empirical research is that it is mostly unidirectional, thus assuming either that policy feedback exists or that opinion representation exists. However, if both mechanisms exist, the assumption of a unidirectional causal influence will lead to biased estimation results. The purpose of this paper is to provide an empirical test of the reciprocal causation between social policy and public opinion. I test this relationship with a two-stage cross-lagged panel model. The project started in November, 2013.

Multilevel Tools - An Ado package for Stata

Together with Katja Möhring. The mlt-package (multilevel tools) is an ado module for Stata. It provides different postestimation commands for hierachical mixed models and some other tools useful for typical tasks in multilevel modelling. We provide an ado to estimate the influence measures (Cook's D and DFBETAs), an ado to obtain Snijders/Bosker and Bryk/Raudenbush R-squared values for multilevel models and an ado to estimate two-stage (slopes as outcomes) models. More ados will be added in the future. Read more about the mlt-package here.

© 2013   |   Alexander Schmidt-Catran   |   All rights reserved  |   Imprint