Protection against Middle East respiratory syndrome coronavirus infection by immunization with genetically engineered liveattenuated viruses.

Abstract

Introduction: Botella et al. (2018) proposed to use the 'non parametric bootstrap' method (Efron & Tibshirani, 1994) to the responses given by an individual to the items of a test in order to create confidence intervals for an individual's true test score for situations in which classical procedures cannot be used. In six databases containing the responses to several psychological scales, two procedures were applied to create the confidence intervals; a classical one, Estimating the True Score (ETS; Gulliksen, 1950), and the Bootstrap of items (BSI). When there was an expected change in the criterion of interest after an intervention and when there was not, the rates of significant change obtained with both procedures were very similar. These results suggested that BSI was a promising solution when other methods could not be applied. However, evidence is needed from different research contexts to assess the performance of BSI.Material and Method: On the basis of the Partial Credit Model (Masters, 1982), a IRT model for polytomous response data, two simulation studies were programmed in R to examine the performance of the BSI technique in comparison to the classical ETS method. In study 1, focused on no change scenarios, we examine the influence of several test features on the width of the BSI confidence interval and its coverage rates of the true score. The factors assessed are the subjects' trait level, skewness of the trait distribution, number of items' categories, number of items, and internal consistency reliability (Cronbach's alpha). Study 2 was carried out to analyze the significant change rates of BSI and ETS given a change in the subjects' trait level.Results: Study 1 shows that the BSI confidence interval is narrower as the subject's trait level gets more extreme and the test internal consistency is higher. The BSI coverage rates of the true score reach appropriate values when the test is made up of at least 20-25 items. Results from study 2 reveal that BSI has lesser statistical power to detect a significant change than ETS as the change in the subject's trait level is bigger. Conclusion:The classical procedure ETS seems to be yet the best option. Nevertheless, the classical procedures for elaborating confidence intervals involve knowing several properties of the test given a fixed sample and, sometimes, these properties are unknown or are not trustworthy. Given the differences in the performance of the methods examined, BSI is a good option to create confidence intervals for an individual's true test score for situations in which classical procedures cannot be used.