<![CDATA[Multiple linear regression is widely used in educational research to examine the statistical association of several predictors with an outcome. Its interpretation, however, depends on appropriate model specification and diagnostic evaluation. This study examined a two-predictor regression model of students’ learning outcomes and evaluated the statistical assumptions preceding model interpretation. A quantitative survey design was conducted at SMA Negeri 1 Rantau Bayur. The sample comprised 40 students selected purposively. Data were collected using Likert-scale questionnaires whose validity and reliability had been assessed and were analyzed using IBM SPSS Statistics. The analysis included descriptive statistics, distributional diagnostics, homoscedasticity, multicollinearity, heteroskedasticity diagnostics, and multiple linear regression. The reported results showed Shapiro–Wilk p values above .05 for X1, X2, and Y; a Levene statistic of 1.244 (p = .272); tolerance of .712 and VIF of 1.404 for both predictors; and no apparent heteroskedastic pattern in the residual scatterplot. The regression model yielded positive coefficients for X1 (B = 0.413, SE = 0.098, p < .001) and X2 (B = 0.287, SE = 0.081, p = .001), with R = .742, R² = .551, and adjusted R² = .526. Thus, the two predictors jointly accounted for 55.1% of the observed variance in learning outcomes in this sample. The study highlights the importance of regression diagnostics while emphasizing that purposive sampling and a single-school sample constrain generalizability and that the observed associations should not be interpreted causally.]]>
