In Global Context: Mental Health Study
Indiana University . Bloomington . Indiana
Finland Sample Design
World Health Survey 2002
– Sampling Plan
This is a stratified cluster sampling, where
clusters are formed from zip code areas (total number: 2865). Primary
stratification is based on municipality data and secondary stratification
is based on zip code areas.
2. Primary stratification
The primary stratification is based on two
these two dimensions six primary level strata are obtained. The
Capital Region (
Since the sample size is 1000 persons and one
starting/sampling point contains six interviews, the total number of
starting/sampling points is 167. Starting points are distributed among
strata in accordance to the population (aged 18+). Thus
the distribution of starting points is as follows:
3. Secondary stratification
second stage stratification groups zip code areas within the first stage
strata in such a way, that the second stage strata are as homogeneous as
possible in terms of socio-economic dimensions. This increases the
accuracy of estimation. In total there are 16 second stage strata. Since
there is not very much information available from zip code areas, it was
decided to use mean income in the capital region and proportion of apartment buildings in other regions as the
stratification variable. At each first stage stratum the number of
starting points was distributed as equally as possible amongst the second
stage strata. From this followed, that within each first stage stratum
each second stage stratum contains approximately similar number of people.
Capital region was divided into four second
number of people and clusters and the sample sizes are in the following
table. The total number of clusters was reduced to 2830, because 35 zip
code areas have no inhabitants.
The number of starting/sampling points inside
each stratum is in a direct relation to the number of inhabitants in each
stratum. Therefore proportional allocation has been used.
Inside each stratum, clusters (zip code areas) have been picked
using PPS-sampling and inside each picked cluster the same sample size (6
interviews) has been used. From this follows, that each person has similar
probability to be picked into the sample – tn=1000/4005817=0,0002496. In
this case estimators are self-weighting.
Inside each sampled cluster (zip code area), a
starting point is chosen randomly.
(zip codes) were sampled with Visual Basic –programmed MS Excel
sample was prepared in cooperation with Dr. Manninen, Professor of
1022 E. Third Street, Bloomington, IN 47405 (812) 855-3841