The issue of nonresponse to surveys is a serious problem in survey research because it reduces the amount of information obtained, creates a significant nuisance for data analysis and may introduce bias into the survey results by flawing the representativeness of the target population under investigation. This paper examines item-nonresponse and "don't know" responses in the General Social Survey of Canada of 1985 and comments specifically on the different effects of personal and telephone interviewing technique on rate of item omission. The effect of sensitivity of the question, and position of the item in the questionnaire on nonresponse are also examined in an exploratory descriptive analysis. Results of an analysis of variance and multiple regression/correlation indicate that both sensitivity and position are factors influencing item omission. In addition, demographic characteristics were found to be significantly and differentially correlated with item omission and "don't know" responses by topic area for the survey under investigation. The results of this study further demonstrate that telephone interviewing significantly increases the rate of item omission, especially for questions on social support.

No description available.

Although the primary role of the computer has been in processing and analysis of survey data, it has increasingly been used in data collection. Computer surveys are not exempt from a common problem: some refuse to participate. Many researchers and practitioners indicate the refusal problem is less for computer surveys, perhaps due to the novelty of the method. What has not been investigated is the refusal problem when on-line surveys are no longer novel. This research study examines the use of one form of computer-assisted data collection, the electronic or on-line survey, as an organizational research tool. The study utilized historical response data and administered an on-line survey to individuals known to be cooperative or uncooperative in other on-line surveys. It investigated nonresponse bias and response effects of typical responders, periodic participants, and typical refusers within a sample of corporate employees in a computer-interactive interviewing environment utilizing on-line surveys. The items measured included: participation, respondent characteristics, response speed, interview length, perceived versus actual interview length, quantity of data, item nonresponse, item response bias, consistency of response, extremity of response, and early and late response. It also evaluated factors reported as important when deciding to participate, preferred data collection method, and preferred time of display. Past participation, attitudes toward on-line organizational surveys, response burden, and response error were assessed. The overall completion rate of 55.7% was achieved in this study. All effort was made to encourage cooperation of all groups, including an invitation to participate, token, on-line pre-notification, 800 number support, two on-line reminders, support of temporary exit, and a paper follow-up survey. A significant difference in the participation of the three groups was found. Only three demographic variables were found to be significant. No significant differences were found in speed of response, interview length, quantity, item nonresponse, item response bias, consistency, and extremity. Significant differences were found in the perceived and actual times to complete the survey.

Non-response is an error common to most surveys. In this dissertation, the error of non-response is described in terms of its sources and its contribution to the Mean Square Error of survey estimates. Various response and completion rates are defined. Techniques are examined that can be used to identify the extent of nonresponse bias in surveys. Methods to identify auxiliary variables for use in nonresponse adjustment procedures are described. Strategies for dealing with nonresponse are classified into two types, namely preventive strategies and post hoc adjustments of data. Preventive strategies discussed include the use of call-backs and follow-ups and the selection of a probability sub-sample of non-respondents for intensive follow-ups. Post hoc adjustments discussed include population and sample weighting adjustments and raking ratio estimation to compensate for unit non-response as well as various imputation methods to compensate for item non-response. / Mathematical Sciences / M. Com. (Statistics)

Non-response is an error common to most surveys. In this dissertation, the error of non-response is described in terms of its sources and its contribution to the Mean Square Error of survey estimates. Various response and completion rates are defined. Techniques are examined that can be used to identify the extent of nonresponse bias in surveys. Methods to identify auxiliary variables for use in nonresponse adjustment procedures are described. Strategies for dealing with nonresponse are classified into two types, namely preventive strategies and post hoc adjustments of data. Preventive strategies discussed include the use of call-backs and follow-ups and the selection of a probability sub-sample of non-respondents for intensive follow-ups. Post hoc adjustments discussed include population and sample weighting adjustments and raking ratio estimation to compensate for unit non-response as well as various imputation methods to compensate for item non-response. / Mathematical Sciences / M. Com. (Statistics)