Validity and reliability of the Survey of Attitudes toward Statistics (SATS) instrument

Abstract

In social science research, statistics play an imperative role. However, social science students’ attitude towards statistics is not widely investigated. This cross-sectional survey validated the Survey of Attitude towards Statistics 28 (SATS-28) instrument, which contains 27 items, in the context of 143 graduate students. Confirmatory Factor Analysis (CFA) was used to determine the compatibility of the hypothetical measurement model with the data obtained and to determine the model’s construct reliability. All factors were successfully replicated in the Malaysian sample. From the initial model, 13 items with low factor loading index were dropped to achieve a good fit, leaving the final model with 14 items within four constructs, which are Affect, Cognitive Competence, Difficulty, and Value. The final model resulted in a lower chi-square (107.455) and AIC (203.455) values, while a higher PNFI value (0.599). Other than that, the cut off value of CFI (0.956) was increased. The RMSEA value (0.050) also fulfilled the suggested threshold of <0.08. Findings also revealed that Construct Reliability (CR) of the final model was acceptable, which ranges from 0.70 to 0.80. The findings implicate that the model is valid and suitable to evaluate Malaysian student’s attitudes towards statistics. It is suggested that similar studies be done on other extended models of the SAT's instrument.