Research Question

The research question was: Does educational attainment and political party affiliation predict attitudes toward marijuana legalization? Using the General Social Survey dataset, I selected "Should marijuana be made legal?" as the dependent variable, "RS highest degree" as a continuous predictor, and "political party affiliation" as a categorical predictor, which SPSS dummy coded. Dummy coding allows categorical variables to be represented in regression models by creating binary comparisons against a reference category, enabling a more precise interpretation of group differences (Warner, 2013).

Interpretation of Coefficients

The results indicated that the overall regression model was statistically significant,F (2,372)=9.75, p<.001F (2, 372) = 9.75, p < .001F(2,372)=9.75,p<.001, and explained about 5% of the variance in support for marijuana legalization (R2=.05R^2 = .05R2=.05). Education had a negative coefficient (b=0.051, p=.014b = -0.051, p = .014b=0.051,p=.014), suggesting that respondents with higher educational attainment were slightly less supportive of marijuana legalization when controlling for political party affiliation. Although this effect was modest, it shows a downward trend in support as education increases. Political party affiliation, dummy coded in the model, was a significant predictor (b=0.047, p<.001b = 0.047, p < .001b=0.047,p<.001), meaning that political identity influenced marijuana attitudes above and beyond the effect of education. As Warner (2013) emphasizes, dummy variable coefficients represent the average difference between the reference group and the coded group, holding other variables constant. This finding highlights the role of political identity in shaping opinions about drug policy.

Regression Diagnostics

To evaluate whether the regression assumptions were met, I examined several diagnostics. The relationships between predictors and the dependent variable appeared linear, and residuals had a mean near zero and a reasonably symmetric distribution, suggesting approximate normality. Homoscedasticity was supported by the relatively stable residual standard deviation (.481), though further inspection with residual plots could provide additional confirmation. Multicollinearity was not a concern, as tolerance values were 1.00 and variance inflation factors (VIFs) were also 1.00. Additionally, the Durbin-Watson statistic (1.914) was close to the ideal value of 2.0, indicating independence of residuals. Taken together, the model met the major assumptions of multiple regression (Wagner, 2019; Warner, 2013). Overall, the regression model supports the conclusion that political party affiliation and education contribute to differences in attitudes toward marijuana legalization, though the model explained only 5% of the variance. This aligns with Warner's (2013) caution that statistical significance does not necessarily imply strong predictive power. Other social, cultural, or demographic factors may be stronger predictors and could be included in future analyses to improve the explanatory value of the model.

References

Wagner, W. E. (2020).Using IBM SPSS Statistics for Research Methods and Social Science Statistics(7th ed.). Sage Publications, Inc.

Warner, R. M. (2013).Applied statistics: from bivariate through multivariate techniques(2nd ed.). Sage Publications.

This is a colleague post above.

Hello Tutors,

Respond to at least one of your colleagues' posts and provide a constructive comment on their assessment of diagnostics.

1. Were all assumptions tested for?
2. Are there some violations that the model might be robust against? Why or why not?
3. Explain and provide any additional resources (i.e., web links, articles, etc.) to provide your colleague with addressing diagnostic issues.

Thanks Tutors.