Confirmatory Factor Analysis of the Survey of Attitudes Toward Statistics (SATS)

Background and Purpose

            The Survey of Attitudes towards Statistics (SATS) is a Likert-type instrument with 36 statements ranging from strongly disagrees to strongly agree. The early version of the SATS consisted of six subscales: Affect, Cognitive Competence, Difficulty, and Value, Interest, & Effort. Few psychometric studies have been conducted with diverse college students. An exploratory factor analysis found a more parsimonious solution suggesting that the SATS-28 may only pertain to two underlying factors, namely (1) the value of statistics as a tool in students’ respective fields of study and (2) different aspects of how a student will perform in a statistics course. A confirmatory analysis suggested that the SATS-36 can be improved by removing some poorly functioning items. The purposes of current study are 1) to examine the overall psychometric property, & 2) to propose for the most refined version.   

 

Methods

            A cross-sectional survey data was collected from 197 social work students who were enrolled at an urban university in the south.  A majority of sample is female (90.3%), African Americans (86.7%). The mean age of this sample is 32.66 years old and over 53% of participants were from graduate programs. Data analysis consisted of a series of statistical techniques including confirmatory factor analysis (CFA). With the SPSS 21.0, item analyses were conducted to verify the scale’s reliability. CFA involved a structural equation model (SEM) using the PRELIS 2.8 and LISREL 8.8.

 

Results

The original reliability study reported that coefficient alpha values are varied somewhat by component: Alpha values are from acceptable to excellent for Affect =.793; Cognitive Competence = .797; Value = .802; Difficulty = .726; and Interest = .943. Item statistics detected several poorly functioning items which lowered the internal consistency of sub-scale. The original five-factor model with 32 item does not fit well (e.g., CFI = .87, RFI = .80).  To identify the most rigorous model, the items with less than R² =.40 (factor loading to factor) are removed. Considering theoretical and practical provision, the five-factor inter-correlated model with 18 items is the best fitted model (e.g., CFI = .97, IFI =.97) based on the fit statistics.  

 

Implications

The findings of this study provide strong evidences for the construct validity and reliability of the refined SATS-18 scale, which will better measure diverse attitudes toward learning statistics. An understanding of these attitudes with more efficient instrument is necessary to help instructors facilitate the learning of statistics for social work students, by enabling them to create more positive attitudes toward statistics.  

Most of social work students already had negative feelings about themselves in relation to the learning of quantitative issues including disbelief about the relevance, therefore, lack of statistics for their future career. In turn, as social work students would perceive statistics and research skill more relevant to their employment, they would take more advanced courses beyond those of minimum requirements. This study further suggests that effective application of the EBP to their job plans requires more research proficiency as well as basic level of statistical knowledge.