W pracy podjęto rozważania nad efektywnością w badaniach rolniczych trzech schematów losowania - prostego ze zwracaniem, warstwowego oraz z prawdopodobieństwami wyboru proporcjonalnymi do cechy dodatkowej ze zwracaniem. Rozważania te wsparte zostały badaniem symulacyjnym z wykorzystaniem danych rolniczych, pochodzących z Powszechnego Spisu Rolnego 2002. Z badań wynika, że najefektywniejszym schematem spośród trzech badanych jest losowanie warstwowe
EN
An optimal sampling scheme is the one in which inclusion probabilities are proportional to values of a survey variable. Of course such a design does not exist, because we do not know the variable values before a survey. Although, if we know values of auxiliary variable that is strongly correlated with the survey variable, we can use them to calculate the inclusion probabilities, that are proportional to the values of such variable. The results of such procedure, i.e. a precision of estimation, depend on a coefficient of correlation between a survey and auxiliary variable. Unfortunately, such schemes are rather complicated, and very often do not exist easy to implement unbiased variance estimators. That is why we can decide to use a pps sampling scheme, i. e. with replacement probability proportional-to-size sampling. Although, in agriculture surveys carried out in Central Statistical Office the most popular sampling scheme is stratified sampling. The paper contains the considerations on efficiency of three sampling schemes, i.e. the simple random sampling without replacement, stratified sampling and pps sampling. The author carried out a simulation study using data from Agricultural Census 2002.