SAMPLING DESIGN AND ESTIMATION FOR HIS AT STATISTICS LITHUANIA

Transcription

1 SAMPLING DESIGN AND ESTIMATION FOR HIS AT STATISTICS LITHUANIA Danutė Krapavickaitė Chief statistician Meeting of the Technical Group Health Interview Survey (HIS) Statistics

2 Outline 1. Main survey information 2. Substitution of sampled elements 3. Non-response and calibration of the design weights 4. Income imputation 5. Difficulties overcomed Meeting of the Technical Group Health Interview Survey (HIS) Statistics 2

3 Main information about HIS at Statistics Lithuania Carried out on September-November 2014 Survey population Sample size Number of respondents sampled units (88 %) substituted units 643 (12 %) Nonrespondents (26 %) Reasons for non-response refusal 855 (48 %) non-con 546 (30 %) imperfections of the sampling frame - unit resides other address 107 (6 %) other 394 (22 %) Household income indicated (91 %) missed 485 ( 9 %) Meeting of the Technical Group Health Interview Survey (HIS) Statistics 3

4 Reason for substitution imperfections of the sampling frame If the interviewer finds out that the individual selected to the sample does not live at the address indicated, but lives in other address because of household decomposition address indicated is not the main place of residence of the unit another household is living at the address selected (house is rented or sold) then it is considered that the address is selected to the sample, such a sampled unit is dropped out of the sample unit is substituted by the member of the actually living household selected by the rule of forthcoming birthday Reasons: no possibility to follow up the respondents the way to include into the survey individuals registered in a wrong way the way to increase response rate Meeting of the Technical Group Health Interview Survey (HIS) Statistics 4

5 Sampling design and estimation method Sampling design simple random sampling of individuals. Data collection by the interviewers. Estimation method calibration of the design weights by the age_group (<25, 25_34, 35-44, 45-54, 55-64, 65-74, 75-84, >84) x sex urban living place (17 groups of respondents totally) Definition. Calibrated weights are the weights which satisfy conditions: Are as close as possible to the design weights in the sense of some metric, Estimate the totals of the selected auxiliary variables exactly In our case the exactly estimated totals are population size by 17 groups: age_group x sex and urban living place groups. Aim of the calibration: to take into account frame imperfections, to correct for unlucky selected sample to take into account nonresponse Meeting of the Technical Group Health Interview Survey (HIS) Statistics 5

6 Estimates of the population group sizes using respondents when nonresponse is not taken into account Meeting of the Technical Group Health Interview Survey (HIS) Statistics 6

7 Estimates of the population group sizes using calibrated weights of the respondents Meeting of the Technical Group Health Interview Survey (HIS) Statistics 7

8 Analysis of equivalent household member income Equivalence income of a household member depends on his/her age living place (city, urban, rural area) household size Imputation of absent income Income homogeneity groups constructed by cross-classification of 5 age groups: <25, 25_34, 35-44, 45-54, >54 territory: cities, other urban areas, rural areas equivalent household size: 1, 2, >=3 45 groups totally Meeting of the Technical Group Health Interview Survey (HIS) Statistics 8

9 Meeting of the Technical Group Health Interview Survey (HIS) Statistics 9

10 Imputed income for a household member Income is simulated by normal distribution - group income mean estimate - group income variance estimate Meeting of the Technical Group Health Interview Survey (HIS) Statistics 10

11 Some difficulties with respect to the data aggregation Mental health. No rules of data aggregation included into the HIS Wave 2 Methodological manual. The solution is found on the internet in the research paper Kurt Kroenke, Tara W. Strine, Robert L. Spitzer, Janet B. W. Williams, Joyce T. Berry, Ali H.Mokdad, The PHQ-8 as a measure of current depression in the general population, Journal of Affective Disorders, (2008), doi: /j.jad Social support. No rules of data aggregation included into the HIS Wave 2 Methodological manual. Oslo-3 Social support scale is found in European Union Public Health Information System on the internet Meeting of the Technical Group Health Interview Survey (HIS) Statistics 11