| Sumario: | The article describes two statistical techniques for measuring error that result due to variance among interviewers. That is, it illustrates these techniques for measuring interviewer and question objectivity. It is stated that the problem is to compare the variance found within the total population with that found among the sub-groups. The information gathered by a particular interviewer might be considered as a sub-group. The first method is to measure whether the variance among the averages of the subgroups exists for a given characteristic. It is what one would expect if these sub-groups were chosen at random from the total population. In the second statistical method of measuring interviewer objectivity, it is necessary to set up an analysis of variance table so that the contributions to variance of the subgrouping and of the differences within sub-groups might be compared. But, in practical sense, the second method is prevented because it demands replicated interviewer assignments. So, authors find the first method to be more successful. It is suggested that more data of the nature obtained be assembled in order to build up a library of acceptable and unacceptable questionnaire practices.
|
|---|