help for imputerasch                                       Jean-Benoit Hardouin

Imputation of missing binary variables by a Rasch model

imputerasch varlist [, prefix(string) norandom saveproba(string) nbiteration(#) details max(#)]

varlist is a list of two or more existing dichotomous variables.


imputerasch imputes missing binary data by a Rasch model. The parameters of the Rasch model are estimated on complete data, then the missing data are imputed from the estimated probability for each individual to response to each item. By default, the imputed value is a result of a random draw within a Bernouilli random variable with this probability used like parameter, but it is possible to affect more deterministically the value of the missing data (0 if p<0.5 and 1 if p>=.5) with the norandom option. An iterative procedure can be run in a second time by estimating parameters of the Rasch model on existing and imputing data, and by eventually correcting missing data at each step (see the nbiteration option). This procedure is stopped as soon the allowed maximal number of iterations is attained, or as soon the imputed values are stable.


prefix. The former variables (with missing data) are keeped. New variables are created by imputing new values to missing data. The name of these new variables are the names of the former variables preeceded by the prefix defined in this option. By default, this prefix is "imp".

norandom avoids to randomly draw the value of imputation (by default). A deterministic process is used : if the expected probability is <0.5, imputed value is 0, else imputed value is 1 (the old name of this option, nobinomial, continues to run).

saveproba allows saving the expected probability in variables whose the names begin by the string defined in this option.

nbiteration realizes an iterative procedure which is stopped as soon as the maximal number of iterations is attained, or as soon as the imputed data are stable.

details gives details on the imputation.

max allows imputing missing values only for individuals with a maximal number of missing values defined with this option.


. imputerasch item*

. imputerasch item*, norandom saveproba(p) prefix(dataimputed) max(4)

. imputerasch item1-item5, nbiteration(5) details


Jean-Benoit Hardouin, PhD, assistant professor EA 4275 "Biostatistics, Clinical Research and Subjective Measures in Health Sciences" University of Nantes - Faculty of Pharmaceutical Sciences 1, rue Gaston Veil - BP 53508 44035 Nantes Cedex 1 - FRANCE Email: jean-benoit.hardouin@univ-nantes.fr Websites AnaQol and FreeIRT