For the sample selection, 14 strata were built combining areas (Metropolitan Lima and Callao, Urban and Rural Rest) and the district poverty quintile. In each stratum, a multiple of 4 clusters was selected by random sampling without replacement and with probability proportional to the total size of inhabitants. 79 of the 80 selected clusters were executed, each of approximately 50-100 households where eight teams worked for 13 weeks in the last quarter of 2015.
In order to reduce systematic errors, in each area they were randomly permuted to assign them to the teams and weeks. In each conglomerate, the random selection of the assigned day was prepared for each of the 10 children of the sample quota. Field teams performed a quick enumeration of the total number of eligible children. When the total was greater than 10, the first 10 were selected according to a pre-selected and different random sequence for each cluster. When the total was less than 10, we worked with all of them and it was not necessary to look for replacements.
The sample size was estimated by stratified and multistage random sampling in three domains (Metropolitan, Urban and Rural Lima). The sampling frame consisted of information on population and housing from the 2007 National Censuses: XI on Population and VI on Housing, available at the National Institute of Statistics and Informatics (INEI).
