Bookmark and Share

Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: st: svyset and complex design

From   Steve Samuels <>
Subject   Re: st: svyset and complex design
Date   Tue, 24 Jan 2012 19:09:45 -0500


I am assuming that sampling is without replacement (WOR) at every stage, since you don't say otherwise.

1. You don't "tell" Stata about PPS sampling.  That information is incorporated in the probability weights. 

2. The fpc specifications should be used only estimating descriptive statistics about the population.

3. If you intend  analytic investigations (e.g. tests and assessment of odds ratios, correlations, regression coefficients), then these analyses, write another -svyset- statement that omits all fpcs and stages beyond the first. For references, see:

You might also consider non-response and post-stratification adjustments to the weights.  

I am curious:  Your second stage design implies that the presence or absence of migrants was known for every household in the sampled districts. How was this information obtained?


On Jan 24, 2012, at 8:47 AM, BRINGE Arnaud wrote:


I want to specify my sample design which is following :

First stage : Sampling of district proportional to size
Second stage : Sampling of households, stratified according to the presence of migrants in the households (Yes/No), unequal probiblities
Third stage : Sampling of Individuals, stratified according to his migrant status (Migrant return, Migrant conjoint, No migrant). Up to 2 migrant return and migrant conjoint, and 1 random No migrant if present in the household.

I have two questions :
1) How to specify the first level is sampled proportional to size (PPS)  ?
2) I tried to specify this design with svyset instruction. Is it correct ?

svyset num_dr [pweight=weight_ctry], strata(Strata_area) fpc(t) || n_menage, strata(Strata_hh) ///
fpc(pmig) singleunit(centered) || _n, strata(statu_mig) fpc(pmig_i)

with num_dr= District identification
weight_ctry=sample weighting
t= Probability of sampling districts
n_menage=Household identification
Strata_hh = Presence of a migrant in the household
statu_mig= Migrant status of individual
pmig= Sampling probability for one household
pmig_i = Sampling probability for one individual


*   For searches and help try:

*   For searches and help try:

© Copyright 1996–2018 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   Site index