Path: janda.org/c10 > Syllabus > Topics and Readings > Modeling Relationships >
More on Multiple Regression in Research
 Multiple Regression in Africa Other examples of multiple regression in research: Predicting African coups Wells: "The Coup d'Etat in Theory and Practice: Independent Black Africa in the 1960s" Wells notes that most scholars said that it was impossible to specify variables which would distinguish countries that had coups in the 1960s from those that did not.  He built a dependent variable -- "coup activity score" -- by assigning 10 points for a successful coup, 3 for an overt attempt that failed, and 1 point for an uncovered plot. He identified 15 likely causal variables of coup activity, conceptually grouped into 8 socioecnomic factors 7 military factors  The simple correlations predicting to coup activity for 31 nations averaged about .10 for the socioeconomic variables and .30 for the military variables  Used multiple regression to combine these factors The 8 socioeconomic variables yielded an R of .41, R2 = .16 The 7 military variables gave an R of .48, R2 = .23 Together, they gave an R of .75, R2 = .56  Conclusion was that a model can be constructed that explains a substantial proportion of coup activity  Problems with the analysis -- done over a decade ago Erroneous claim: "The [beta] coefficients are interpreted as unstandardized partial correlations and indicate the contribution of each variable to the total explained variance in the dependent variable." (pp.883-884) What is wrong about this statement? Table 4, which reports the "[beta] coefficients and multiple correlations" provides insufficient information to judge the analysis -- e.g., no idea of level of significance set for the coefficients The combined analysis, which explained 56% of the variance, used 15 variables for 31 cases -- what is the problem here? Jackman: "The Predictability of Coups d'Etat: A model with African Data" Jackman's article is more sophisticated than Wells, and you are not expected to understand it all. Jackman explained 79% of the variance in coups for 29 countries with only 7 variables The variables were entirely different from Wells' Social mobilization Pluralism (size of ethnic group) Party strength Turnout In general, Jackman's work was more theoretically oriented and better executed than Wells'. These readings were chosen to illustrate the point that to understand and analyze the current literature in political science, one needs to know something about the methodology employed.