Introduces overall content of the KLoWF,
survey areas and methods
Needs for weights
For a longitudinal survey, not only cross-sectional weights but also longitudinal weights are needed for the differences in probabilities of selection, responses and tracking.
As KLoWF has households and women aged from 19 to 64 living in the households as samples, household weights and female individual weights are needed respectively.
Only cross-sectional weights were calculated for the first wave KLoWF data because the first wave data were cross-sectional. With the progress of the subsequent wave surveys, however, longitudinal weights will be applied at the same time.
Kinds of weights
Household weight: This weight can be used when the analysis unit is a household.
Enumerated household member weight: This weight can be used when the analysis unit is an eligible female household member in a household
Responding household member weight: This weight can be used when the analysis unit is an eligible female household member who responded in a household
Selection probability
The probability of each household and household member being selected as a sample and responding can be expressed as follows
Selection probability of sample households = (ED selection probability)×(probability of eligible households being selected as sample households in an ED and responding to the questionnaire for households)
Selection probability of sample enumerated household members= (ED selection probability)×(probability of eligible households being selected as sample households in an ED and responding to the questionnaire for households)
Selection probability of sample responding household members= (ED selection probability)×(probability of eligible households being selected as sample households in an ED and of eligible household members responding to the questionnaire for individuals)
In all cases, the ED selection probability, which comes first in all these formulas, is proportional to the square root of the total number of households in the region to which each household belongs.
Probability of eligible households being selected as sample households in an ED and responding to the questionnaire for households = (total number of eligible households contacted/ total number of eligible households)×(probability of eligible households contacted responding to the questionnaire for households)
Probability of eligible households being selected as sample households in an ED and of eligible household members responding to the questionnaire for individuals= (total number of eligible households contacted/ total number of eligible households)×(probability of eligible household members from eligible households contacted responding to the questionnaire for individuals)
As it was impossible to calculate the systematic response rate in all three cases, this study calculated the basic weights assuming that there was no systematic difference if households existed in each ED.
Basic weights for households(XW1_H) = (ED selection probability)×(eligible responding households/eligible households)
Basic weights for enumerated household members (WX1_PE) = (ED selection probability)×(eligible responding households/eligible households)
Basic weights for responding household members(WX1_PE) = (ED selection probability)×(eligible responding households/eligible households)
For details and procedures, see <2008 KLoWF: Project Report and the First-wave Descriptive Analysis Report> and <KLoWF Brief, V. 4>.