Page 201 - Contributed Paper Session (CPS) - Volume 1
P. 201

CPS1278 Desiree R. et al.
                The proposed sampling design uses a stratified single-stage design with
            the establishments treated as sampling units. The data used for this study is
            the  list  of  establishments  from  the  2018  Integrated  Survey  of  Labor  and
            Employment (ISLE). Meanwhile, the 18 PSIC Sections were considered as the
            primary strata or domains of the survey. For each domain, four strata were
            formed which are defined in detail in Table 1.

                             Table 1. Characteristics of Strata and Sampling Strategies
              Stratum                 Characteristic                 Sampling Strategy
                 0     Very small establishments whose aggregate TE     Take none
                       contributes  at  most  5%  of  TE  of  all
                       establishments.
                 1     Smallest stratum formed by the LH algorithm      Take none
                 2     Medium stratum formed by the LH algorithm        Take none
                 3     Largest stratum formed by the LH algorithm        Take all

                For stratum 0, no establishments will be taken as these only contributes
            very small amount to the overall total employment (TE) of the establishments.
            Meanwhile, for strata 1, 2, and 3 the stratum boundaries will be obtained using
            the LH algorithm.
                The optimal sample size will also be obtained using LH algorithm for strata
            2 and 3 while all establishments falling within the boundaries of stratum 3 will
            be taken as samples.
                As adapted from the 2011 Survey of Enterprises in the Philippines5, the
            establishments were sorted according to their region and was followed by
            systematic  selection  of  samples  in  order  to  ensure  that  samples  were
            distributed  across  each  region.  This  method  would  also  induce  implicit
            stratification among the regions.
                As the dataset used is a survey data, the method of obtaining samples
            using LH was simulated for 200 times. This is to show that the CV for each of
            the samples obtained using LH would approach the expected target CV.

            3.  Results
                In  order  to  obtain  the  optimal  sample  size  and  boundaries  for  each
            stratum, cutoff sampling together with LH algorithm was applied. The resulting
            sample size excludes the establishments whose aggregate TE contributes at
            most 5% of TE for all establishments or those belonging to stratum 0. Hence,
            the samples consist of establishments belonging to strata 1, 2 and 3 only. The
            resulting sample size, the percentage of samples and the corresponding CV of
            TE for each section is shown in Table 2. Meanwhile to further inspect the CV
            of TE for the samples obtained, the CV for each subsection was obtained and
            plotted per section as shown in Figure 1.


                                                               190 | I S I   W S C   2 0 1 9
   196   197   198   199   200   201   202   203   204   205   206