A Multiple Regression Model of Statistical Reasoning: A Malaysian Context
12 Pages Posted: 11 Jan 2017
Date Written: October 30, 2016
There is a noticeable slide in Malaysian mathematics achievement reported in the last three TIMSS studies and in particular in the area of Data and Chance. This is actually not only unique to Malaysia but also in many other countries in Asia and Africa. Recent studies have shown the influence of higher order thinking skills like reasoning and decision making on statistics achievement. Chan, Ismail and Sumintono (2014) found that statistical reasoning among Malaysian secondary school students to be poor. The purpose of this study is to determine the influence of language, and misconception on statistical reasoning using a sample size of 374 Diploma of Science students from a campus of a large Malaysian public university. A quantitative research design was employed as the objective of this study was to measure the strength and direction of the effect. The flexibility and power to analyze complex multivariate relationships concurrently are possible using a multivariate linear regression approach. The research procedure included a pilot study to determine the feasibility of the procedure and suitability of the adapted Statistical Reasoning Assessment (SRA) to the population of interest. A survey form was used to collect both primary and secondary data. The form comprised of items to collect respondent profile information, grades from relevant courses they took previously and self-reported grades of their mathematical achievement and language proficiency in the public examinations. The findings showed that students did not do well in statistical reasoning (SR) and had a substantially high level of misconception (MC) about statistics. SR (M = 38.17, SD = 13.83) and MC (M = 34.44, SD = 11.56). Language mastery (ENG) was found to be above average, (M = 3.26, SD = .73). The regression coefficients indicated that Language mastery (ENG), and Misconception (MC) significantly predicted Statistical Reasoning (SR). The best model generated was . The coefficient of determination for the regression model was R2 = 0.309 indicating that ENG and MC alone explained 30.90% of the total variance. Squared semi-partial correlation (sr2) informs us of the unique variance explained by each of the variable. sr2 for ENG is given by (.186 X .186 = .035) while MC is calculated by using (-.493 X -.493= .243). These indices showed that ENG and MC accounted for 3.5% and 24.3% respectively of the variances. This paper concludes with a discussion on the pertinent issues related to the administration of the SRA instrument and recommendation for further research in the field of language, statistical reasoning and misconception.
Keywords: Language, misconception, regression model, statistical reasoning
Suggested Citation: Suggested Citation