There is an increase in the use of national and international large-scale assessment data to describe what groups of students know and can do. These assessments achieve a broad coverage of the content domain by dividing the pool of items into blocks of items, and administering to each student a subset of the assessment pool. Because students are only measured with a subset of the items, the measurement of individual proficiency is achieved with a substantial amount of error. Under these conditions, traditional approaches to estimating individual proficiency, such as marginal likelihood (MLE), and estimated-a-posteriori (EAP) result in biased estimates of group level statistics, whereas scores in the form of plausible values result in unbiased estimates. Plausible values are proficiency estimates based on student responses to the cognitive items administered in combination with their background information. In this paper we make use of simulated data to show the advantages of using plausible values over traditional point estimates of individual proficiency. We simulate response data under different conditions (number of items) and compare the results under different estimation methods showing the amount of bias for the mean, standard deviation and percentiles of a distribution when plausible values are not used.