Dear imputers, I have the feeling that I've been addressed (by Seppo) in some sense, because we both are involved in DACSEIS, Seppo with SI, me with MI.
I am very enthusiastic about MI, but please, not fanatic!!! I hate fanatism, in any sense. My enthusiasm is based on a growing experience with MI first in simulation setups and now with large data sets, even in complex surveys, even with establishment panel data etc. In my opinion, it is the most flexible technique allowing to correct for nonresponse (item- but, in some cases, also unit-nr), useful with sophisticated econometric models, even with propensity score matching or with split questionnaire designs or for disclosure control. The chained equations provide very promising and flexible imputation models for large scale data sets with variables of different scale. I agree that multiply imputing complex data is not an easy task, but the same holds for a good hot-deck imputation. And if you can do it once, why not multiply? Then, in general, the analyst can proceed with standard complete-data inference. The possible separation of imputer and analyst is another nice feature of MI. Like Rod, I cannot understand why NSIs will not, shall not, or cannot use MI when providing public use data sets. In the DACSEIS project right this moment, a rather "neutral" person, Ralf, is computing different routines in a simulation setup based on real German microcensus data, comparing hot-deck imputation and variance correction, resampling methods and MI concerning typical survey estimates, i.e. the HTE and GREG. Of course, every method has its merits and demerits here, the results will be presented at the European Conference on Quality and Methodology in Official Statistics in Mainz, Germany, May 24-26, if interested see http://q2004.destatis.de/cp_1.htm We will also present some stand alone software then, allowing e.g. to split questionnaires and to multiply impute large scale data sets with flexible chained equations. Hopefully nobody is annoyed by my short advertisement... but, please, don't call us fanatic, though ;-) Multiple best wishes from Franconia in Germany, Susanne --------------------------------------------------------------------- PD Dr. Susanne R?ssler Department of Statistics and Econometrics Faculty of Business Administration, Economics and Social Sciences Friedrich-Alexander-University Erlangen-Nuremberg Lange Gasse 20 D-90403 Nuremberg Tel: +49-911-5302-276 Fax:+49-911-5302-277 email: [email protected]
