Wardani Rahayu^1,, Dogol Harjono²

The purposes of this study were to develop instrument of high school students’ affection towards mathematics. It also compared estimation of multidimensional reliability coefficients of the instrument, i.e. McDonald reliability and Maximum reliability coefficient. The research used multi-stage random sampling technique. Conceptually, construct of the affection instrument consists of 5 dimensions, namely: attitude, interest, self-concept, values, and morals. Inter-rater reliability coefficient which were obtained for the five dimensions move from 0.73 to 0.80 and produce 86 selected items. These results were achieved from the items which have been selected through expert study and panelists’ validation. Empirically, first and second trials (try out) testing which used confirmatory factor analysis (CFA) with Maximum Likelihood method (ML) recieved loading factor value above 0,3. In the first and second trials, the multidimensional reliability coefficient calculation showed that the value of McDonald and Maximum reliability coefficients were more than 0.9, which means that the level of reliability of the affection instrument towards mathematics of high school students classified in a very high level. Hypothesis testing of McDonald and Maximum reliability coefficient used t-paired test on bootstrapping methods samples. The result is McDonald reliability coefficient is lower than Maximum reliability coefficient.

instrument development, affection towards mathematics, McDonald reliability coefficient, Maximum reliability coefficient