In stratified sampling there are designs with one unit selected per stratum or when one is willing to make estimations on unplanned domains with one unit in some strata. In these cases the variance is generally estimated by the collapsed strata method, which requires identification of the strata to be collapsed previous to variance estimation. This can be quite complex in surveys with a lot of strata or variables to estimate. In this article we propose an alternative method by means of two ratio estimators based on strata means obtained from jackknife subsamples, which are easy to compute without collapsing strata. The estimators are biased and we build the expressions for them, together with their sample estimators. Some examples are given, among them, estimations with data from the Mexican employment survey and poverty measures.
Ratio estimator Resampling Variance estimation Confidence interval Subpopulations Collapsed strata