The Rasch model makes no distributional assumptions of the data under investigation. The unit of measurement in Rasch analysis is the logit, which are interval based. Rasch analysis provides an integrated framework that evaluates if an outcome measure is internally valid and satisfies other requirements for constructing measurement, including the stochastic relationship between persons and items, as mentioned above, and assumptions of local independence, unidimensionality and invariance across groups. Each of these requirements will be explained in brief below. Local independence: To Importazole achieve internal validity a scale must demonstrate local independence, in other words, NSC608001 responses to any given item should only depend on the trait level, and not on responses to previous items. The latter is called response local dependency. With our repeated item design there was a risk that the response to one item was dependent on the response to another item. Therefore, we gave particular emphasis at the outset to the formal test of local dependence. This was examined by examining the residual correlations between items, which should be no more than 0.20 above the average residual correlation. Generally, where items are essentially replicates of existing items, as might be the case in the current design there might be an increase in reliability, and increased variance of person and item estimates. However, the primary goal of this analysis is to examine the scaling properties of the pain VAS, as opposed to validating a scale which has been artificially constructed for this purpose, and thus the concern is with the effect upon the latent estimate, which will be used for comparison with the raw VAS score. Unidimensionality: The Rasch model requires the scale to measure one construct or dimension. This is examined by creating two subsets of items, which are identified by a principal component analysis of the item residuals, with those loading negatively forming one set and those positively loading the second set. Strict unidimensionality is then examined using an independent t-test on the two estimates derived from the subtests for each respondent. If the 95% confidence interval of t-tests include 5%, unidimensionality is supported. Invariance: A scale will consist of items that are easier, and items that are harder to ��achieve�� or ��endorse��.