|
?:comment
|
-
Weighting factor
No total weight usable for international comparism.
NATIONAL weighting procedures/ variables used for calculation of NATIONAL weights:
AT: Age, Sex, Education are used to adjust for non-response bias.
AU: A post-stratification or non-response weight was calculated with the aim of rebalancing the Australian sample so the weighted sample frequencies are equal to the expected frequencies in population in a three-way tabulation by age group (5 groups), sex, and highest level of education (5 levels). The cross tabulation from the Australian population was done using the 2011 Census Table Builder. In the sample not all individuals had complete information on age, sex and year of schooling. However imputations of weight were used to deal so that all individuals were assigned a weight even if they had missing data.
BE: In order to deliver one integrated harmonized data file for Belgium post stratification weights were calculated based upon age (three age groups: 18-39; 40-59; 60+), sex, and geographical classification (NUTS1 Flemish Region, Walloon Region, Brussels Capital Region). These characteristics are known for all sampled units and the population distributions are published by the Belgian Institute for National Statistics (ADS). The age-sex distribution in each region is reproduced by the weights as well as the distribution over the regions. The result of this weighting procedure in combination with the sampling design is that the weights for respondents of the Brussels Capital Region are much smaller than the weights for the respondents of the Flemish and Walloon regions. The use of adequate software is always necessary when analyzing data with weights, but with these data it is vitally important when comparing the regions or analyzing only one region separately.
CH: No weighting.
CL: A weighting procedure is applied in order to correct for distortions in the representativeness of the sample as regards three variables of interest: Gender, Age, (grouped in five categories: 18-24 years, 25-34, 35-44, 45-54, 55 or older) and Urbanity (classification of place of residence as urban or rural). This makes it possible to obtain a sample with characteristics similar to those of the population. The weights are constructed by calculating the quotient between the expected distribution and that observed in the cross between Urbanity, Gender and Age. The expected distribution is obtained from the 2002 census data provided by the National Institute of Statistics. The result of the weighting slightly corrects for problems of under- and over-representation among certain specific groups of the population.
CN: Design weight to adjust for unequal selection probabilities.
CZ: Data was designed using Deming-Stephan algorithm (iterative proportional fitting) with continuous trimming of weights. The resulting range of weights is 0.5-2.0. Weights were designed using most recent dara from the Czech Statistical Office. Weighting factors were age, gender, region, size of community, education and economic activity.
DE: Two separate German samples: the sample for eastern Germany deliberately over-samples the five eastern federal states. If all of Germany is taken as the unit of analysis (rather than the eastern and western states), design weight is necessary: weighting factor for Western Germany: 1.23039027; weighting factor for Eastern Germany: 0.53633957; recoding of the country variable is necessary (see c_sample). The two weighting factors are stored in the weight variable.
DK: No weighting.
EE: Weighting performed based on the national statistical Office 2015 population data; weights are added for gender, age, ethnicity, urban/rural, and regions. Multilevel weighting was performed, considering all pre-defined subgroups. For each of the structural subgroup (such as men aged 15-19) the weights within the group were calculated.
ES: Design weight to adjust for unequal selection probabilities + Weight to adjust for non-response (variables used for calculation of weight: Autonomous Community (ES_REG), Size of habitat)).
FI: The design of the survey was systematic sampling. In order to improve the efficiency of estimation and to reduce bias due to non-response a calibration method was used for the creation of the weight. The following marginal distributions of the population were used: 1) gender (male, female), 2) age classes (15–24, 25–34,…, 65–74), 3) NUTS3 regions with following modifications: the Greater Helsinki Area was dealt as a separate region, 4) type of community (urban - semi-urban - rural).
FR: Post stratification weighting computed on sex, age (4 groups: 18 to 29, 30 to 39, 40 to 54, 55 or more) and occupation (6 groups: Farmers, Tradesmen, Shopkeepers and Business Owners; Managers and Secondary/ University Teachers; Intermediate Professions, White Collar Workers; Blue Collar Workers; Unemployed)
GB-GBN: Selection weights are required because not all the units covered in the survey had the same probability of selection. The weighting reflects the relative selection probabilities of the individual at the three main stages of selection: address, DU and individual. At each stage the selection weights were trimmed to avoid a small number of very high or very low weights in the sample; such weights would inflate standard errors, reducing the precision of the survey estimates and causing the weighted sample to be less efficient. Less than one per cent of the selection weights were trimmed at each stage.
GE: Variables gender and age used for the calculation of weight.
HR: No weighting.
HU: The weighting factor has been built to meet the common frequencies of the four following demographic factors – gender, age, type of settlement and level of education (marginal statistics are derived from the latest Census data of the Central Statistical Office 2011).
IL: No weighting.
IN: Multiple weighing done on following variables: Age, Gender, Social status, Income, Education.
IS: Weighting factor to adjust for non-response bias in terms of IS_REG, AGE an SEX.
JP: No weighting.
LT: No weighting.
LV: Weighting factor constructed according to the following five population characteristics: gender, age, nationality, region and place of settlement.
MX: No weighting.
NO: No weighting.
NZ: Weighting factor adjusts for non-response bias in terms of sex, age, Māori descent, region, rurality, deprivation (neighbourhood deprivation quintiles), and occupation.
PH: Weighting variable is based on people aged 18 and above and region (NCR, Balance Luzon, Visayas, Mindanao). The weight applied was based on the 2016 census population. Computed by area.
PL: Because of household sample, the weighting factor was calculated using two stages. First stage equalized probabilities of being chosen for all respondents. Second stage was an IPF (Iterative Proportional Fitting) algorithm. It includes four variables: 1. sex (male, female), 2. age categories (18-24, 25-34, 35-49, 50-59, 60 and more years), 3. type of the place of living (urban/ rural), 4. voivodship (16 voivodships corresponding to administrative division of Poland). 5 iteration were needed to comply with the population-based distributions with precision of 7 digits after dot. Weighting procedure assumes that after weighting the sample size is equal to the number of completed interviews.
RU: 9 socio-demographic groups based on the variables sex, age and education were used for each region/ strata for the calculation of weight. The weighting procedure is based on the Census 2010.
SE: No weighting.
SI: No weighting.
SK: Weight constructed according to the following population characteristics: sex, age group (18-29, 30-39, 40-49, 50-59, 60 +), education (Lowest formal qualification, Above lowest, Higher secondary completed, University degree), size of community (7 categories: up to 1000, 1001–5000, 5001-20000, 20001-100000, over 100000) and county structure (8 categories).
SR: Weighting factor adjusts for non-response bias in terms of a) district and b) houldehold composition (gender, age, education, main activity, relationship to head of household).
TW: The data was weighted using an iterative proportional raking scheme. For each respondent, sample data were weighted by sex, age, urbanization and education-degree groups. Weights were then generated to match the population characteristics of Taiwan.
US: The US WEIGHT variable takes into consideration a) the sub-sampling of non-respondents, and b) the number of adults in the household. It also essentially maintains the original sample size.
VE: The weight variable was constructed by multiplying the probability of selection by the post-stratification by sex and age.
ZA: Explicit stratification variables: Province, population group and geography type (viz. urban formal, urban informal, tribal and rural formal, including commercial farms). Non-response adjustment = number of drawn HHs per EA (census enumerator area) / number of responding HHs (i.e. where a person 16+ is successfully interviewed) provided that at least 50% of households responded. Otherwise two similar (i.e. in the same explicit stratum) and neighbouring EAs are combined and a combined adjustment factor calculated. Person and household weights are benchmarked, for persons using province, population group, gender and 5 age groups (16-24, 25-34, 35-49, 50-50 and 60 and older) as benchmark variables and for households using province and population group of the respondent in the household. The marginal totals for the benchmark variables are obtained from the applicable midyear estimates as published by Statistics South Africa.
(en)
|
![[This page as RDF]](https://data.gesis.org/gesiskg/static/rdf-icon.gif){kind=icon}