Effect of sample size multidimensional reliability of National Scale Test of Mathematics in DKI Jakarta Science Program High Schools

Abstract

Estimates of internal consistency reliability of the approach are based on data from one form of the imposition of a measure on a group of subjects (single trial administration). There is a relationship between the two sides approach with consistency in estimating the reliability of the mathematical estimation. Difference between the two approaches is the unit of analysis in which the two sides approach compares the first half with the other parts of the test while the internal consistency approach compares each item with every other item on the test. The advantage of using internal consistency approach is, to avoid the problems that are usually caused by a re-test approach and the approach of parallel tests. The shape and nature of the measuring instrument and the many parts that are made will determine the reliability coefficient estimation technique. Among the computational techniques of internal consistency reliability model of a multidimensional approach in this paper discusses the Stratified Alpha, Mosier, and Wang Reliability Coefficient formula. Interpretation of the reliability coefficient is relative. There is no absolute limit which showed the lowest reliability coefficients that must be achieved so that a measurement can be called reliable. Many researchers found that evaluation of the psychometric property measurements using psychometric formulas without verifying the assumptions underlying the formulas can be questioned its reliability. If these assumptions are not met then the Alpha coefficient reliability produced is below the limit value estimate (underestimate). Psychometric experts have warned that most of the formulas contained psychometric assumptions that need to be met in order for these formulas yield accurate information on the measurements made.