Page 43 - Contributed Paper Session (CPS) - Volume 1

CPS798 Quindale E. C. et al.
at the provincial/highly urbanized city (HUC) levels at the end of the year.
Along with this new MS, a new estimation procedure for annualized provincial
estimates was explored. This paper aims to develop an estimation procedure
for the annualized provincial estimates of the LFS.

2. Methodology
The data sets were checked and inspected for completeness and lonely
PSUs. Lonely PSUs were collapsed since at least two responding sample
housing units per PSU were required to have a valid variance. Recomputation
of SSU weights were also performed for these collapsed PSUs. Estimation of
annualized totals and rates at the provincial level was simplified by merging
the provincial data from all replicates for each quarters into one set of data for
the province/HUC to get rid of the individual estimation of replicate levels and
variance. This modifications did not affect the derived estimators for totals and
rates but there is a slight change in variance of totals. The estimation
procedure takes into account the design of the master sample and the
selection of sample housing units for the LFS. The procedure entails two major
steps: computation of survey weights and estimation of population
parameters.
Survey weights are used as raising factor for sample data to produce
estimates for population parameters such as population total, mean and rate.
These weights compensate for the unequal selection probabilities in the
survey design, for nonresponse and for noncoverage. The weights for the LFS
using the 2013 MS design were developed in three stages: computation of
base weights, adjustment of the base weights to take into account unit
nonresponse and population weighting adjustment to make population
counts conform to the population projection.
2.1 Simple Averaging Method
Estimate and variance for the annualized provincial total,
4 4
1 1
̂ ̂
̂ ̂
̂
̂
= ∑ , = 1,2,3,4 (), ( ) = ∑ ,

4 16
=1 =1
where: ≡ annualized estimate of for province , ≡ estimate of for
̂
̂

quarter in province . The rate and its corresponding variance is:
̂
1
̂
̂ ̂
̂ ̂ ̂
̂ ̂
̂ ̂
̂
2
= , ( ) ≈ {( ) + ( ) − 2 ( , )}

̂
̂
2
̂
where: ≡ estimated total of for province .
2.2 Bootstrap Method
For each indicator in each province and in each quarter, bootstrap estimate
and corresponding variance were computed by following the steps below:
1. Draw a simple random sample of size with replacement ( =equal to
the total number of households in the province).
32 | I S I W S C 2 0 1 9