Page 372 - Invited Paper Session (IPS) - Volume 2
P. 372

IPS280 VEUN, Thy et al.



                              Sampling design proposal for Household Survey
                                     in Cambodia with provincial domain
                                                          2
                                Thy Veun , Erniel B. Barrios , Kevin Carl P. Santos 2
                                         1
                  1 Deputy Director, National Institute of Statistics, Ministry of Planning, Phnom Penh, Cambodia
                                         2 University of the Philippines-Diliman

                  Abstract
                  A sampling design for household surveys with provinces as domain is deemed
                  necessary  because  the  estimates  of  households/person  characteristics  at
                  granulated levels rather than the national and regional levels are increasingly
                  becoming important. In this paper, a new sampling design for the household
                  survey  in  Cambodia  with  provincial  domain  is  proposed.  In  this  sampling
                  design,  a  further  stratification  within  each  province  using  the  number  of
                  households in each village is introduced in addition to the sampling designs
                  that  have  been  used  in  the  current  practices  of  the  household  surveys  in
                  Cambodia. In this stratification, each provincial frame, which contains the list
                  of villages, of Cambodia is stratified into five strata, and three methods are
                  used to determine the strata boundaries for each province. These methods are
                  (i)  the  generalized  Lavallée  and  Hidiroglou  (1988)  method  with  Kozak
                  algorithm, (ii) the cumulated root frequency method of Dalenius and Hodges
                  (1959), and (iii) the geometric stratification of Gunning and Horgan (2004).
                  Three levels of required precision (1%, 3%, and 5% CV) are used in conjunction
                  with Neyman allocation to determine the sample size in each stratum and its
                  associated boundaries. Only four among the nine sets of samples from the
                  said stratification with the lowest sample size are finally proposed. The results
                  show that the required sample size in each province and in the whole country
                  is greatly reduced when compared with the case when no further stratification
                  within each province is done. Similarly, the results from the assessment on the
                  performance of the new proposed design using the sample data from the 2014
                  Cambodia Socio-Economics Survey are also found to be acceptable within a
                  precision level.

                  Keywords
                  Stratification;  generalized  Lavallée  and  Hidiroglou  method  with  Kozak
                  algorithm; cumulated square root method; geometric stratification; stratum
                  boundary

                  1.  Introduction
                      A  sampling  design  for  household  surveys  with  provinces  as  domain  is
                  deemed  necessary  because  the  estimates  of  households/person
                  characteristics at granulated levels rather than the national and regional levels
                                                                     359 | I S I   W S C   2 0 1 9
   367   368   369   370   371   372   373   374   375   376   377