Impact of measurement errors on categorical data

Kahiri, James Mwangi Kamau (1995) Impact of measurement errors on categorical data. University of Southampton, Doctoral Thesis.

Abstract

In general, misclassification errors in categorical data affect the results of a survey by introducing biases and a nonsampling variance component on the estimates. The effects of misclassification errors resulting from interviewer-respondent interaction, the recording and coding process are considered under a general sampling and misclassification model. In the general framework, the classification process is regarded as a transition process through steps referred to as phases. Three phases are considered where the first phase involves the interview of the respondent, the second phase involves the recording of interview responses and the third phase involves the coding process.

The biases and the variances emanating from the three transition stages are derived and by simulation process it is shown that under certain conditions, the variance of the estimate is less in the presence of misclassification errors. Also under certain conditions, the resultant effect of errors from various transitions is a reduced bias as well as variance.

The effect of misclassification errors is considered under cluster sampling, assuming a single stage and two stage cluster sampling. Under the assumption that the population size is unknown, the ratio estimator is considered. Finally, the effect of misclassification errors on logistic regression analysis is considered. This involves the effect on the estimation of logistic regression parameters as well as the tests of significance.

This record has no associated files available for download.

Contributors

Download statistics