Abstract

In probabilistic sampling, sample weights are used to construct estimates of means and variances, among other quantities. These weights may include adjustments for non-response, adjustments to known population totals, and calibration with information on auxiliary variables positively correlated with the survey variables of interest; as well as truncation or rounding of sampling weights to integers. The adjustment to integer numbers of the sample weights is carried out in relevant surveys for decision-making such as the ENOE (National Survey of Occupation and Employment) carried out by the INEGI or the monthly survey of the IMEF (Mexican Institute of Finance Executives). This survey is used to construct the manufacturing and non-manufacturing IMEF Indicators. The methodological documents of these surveys do not describe the method used to adjust numbers to integers or the reasons for its use. In this poster it will be shown that the mean estimator under simple random sampling is biased when adjusting the weights to integers . Some examples of these effects will be shown. The effect on systematic sampling will also be discussed.