Applying Rasch modeling to the validity of a control balance scale

Classical test theory approaches to measurement are based upon a number of fairly restrictive assumptions. The nature of these assumptions is such that scales designed to operationalize latent constructs are treated with little regard for the interaction between the human subject and the items that compose the measurement scale. A competing measurement approach, the Rasch model, has been advanced to improve upon the limitations of classical test theory. In this article, the Rasch model is applied to a scale designed to measure Tittle's control balance ratio. Analysis suggests that the original scale and response category options do not meet fundamental measurement. In particular, several items exhibited misfit (and were thus removed from the scale), response category options were reduced from eleven to four, and the scales that form the control balance ratio exhibited very little gender bias. An alternative control balance scale that retains the original items is presented, but this scale no longer summarizes the control balance ratio as one overall score. The implications of these results are addressed.