Conferences

The objective of this work is to study, define and compare alternative sampling frames (population lists) for the representative population coverage as a base for sample selection in internet surveys. In addition, the study aims to provide a methodology for the domain adjustment procedures which enables to use the latest available information. There are several types of sampling frames such as target population frame and available list frame. The target population frames are the ideal frame, which cannot easily be obtained. On the other hand, total population list frames can be either household or individual unit based. The quality of the existing sampling frames can be can be different. The sampling frames may be; undefined, ill defined, partially defined, or perfectly defined.

The target population of the sample information collected from internet surveys (e-mail surveys & web surveys) faces many difficulties in the identification of the possible sampling frame. For voluntary participation, one has to identify the coverage of the target population. Within the available data sources, special adjustments are proposed for the small domains. Some basic variables can be proposed for this purpose. Adjustments can be made for gender breakdown, age groups, and education groups.

In terms of sample selection, the type of access to internet surveys can be based on several limitations. Early attempts was based on the restricted access designs for e-mail survey, or e-mail message followed with a web survey. As an alternative, voluntary participation designs became common for web surveys. In this type of design, observations are obtained through haphazard entry to the web survey questionnaire. Therefore, a probability sample cannot be obtained, due to unknown selection probabilities. Alternative data adjustment procedures for the case of voluntary participation is proposed in this study. Several types of weighted location estimators are proposed on the basis of these designs. For surveys having complex sample designs, a combined ratio mean (proportion, mean, or ratio) or separate ratio mean alternatives can be proposed.

The overall selection probabilities for households can be determined by obtaining household based information from the population information for each domain. This information can be obtained by dividing the total population of each domain by the corresponding average household size for this domain. The overall selection probability of the domains will be their overall sampling fractions. From the previous experience with the variability of the average household sizes in different domains, an alternative approach is also proposed. Due to the unavailability of the census data as Urban and Rural, it was decided to evaluate the available version as City (province centre & district centre) versus Village (sub-district centre & village).
In this study, alternative sampling frames are compared for their population coverage and representation for sample selection in internet surveys. The work also aims to provide a methodology for the domain adjustment procedures which also enables to eliminate the sample selection bias on the basis of the corrected information.