Determination of Implicit Stratification Variables for the Development of 2023 Geo-enabled Mastersample for Household-based surveys in the Philippines
64th ISI World Statistics Congress - Ottawa, Canada
Format: CPS Abstract
Session: CPS 31 - Survey statistics VII
Monday 17 July 5:30 p.m. - 6:30 p.m. (Canada/Eastern)
The main goal of this study is to re - examine if the implicit stratification variables used in the 2013 Master Sample are still adequate for the 2023 Geo-enabled Master Sample, and if not, to develop more up-to-date, efficient, and cost-effective implicit stratification variables to produce reliable household-based survey estimates in the Philippines. Eight (8) implicit variables that are tested in thirty – nine (39) combinations or two hundred thirty – four (234) permutations of implicit stratification variables were simulated to determine which set of implicit stratification variables produced the lowest coefficient of variation (CV), standard error (SE), design effect (deff), and relative bias (RB).
Comparing the sets of implicit stratification variables used in the 2013 Master Sample with the sets of implicit stratification variables with the lowest average CV, average SE, average deff, and average RB across the ten variables of interest, the set of implicit stratification variables with the most instances of having the best results is agricultural score plus wealthproxyB plus geolocation when using the matched 2018 LFS-FIES and 2013 Master Sample. But with the most recent 2021 LFS-FIES, further investigation is conducted to validate the initial results obtained.