Statistical Evidence in Sociological Research - 2/9/09

Statistical Evidence in Sociological Research CREATEDATE 2/9/09 2:59 PM Word of the Day Hypothesis ? a supposition or conjecture to put forth to account for known facts Why did attendance at private academies rise so dramatically in Mississippi from 1969 to 1971? The Logic of Sociological Research Sociological research typically begins with an explanatory question Why did capitalism arise in the west Why do people commit suicide Why did social revolutions occur in France, Russia, and china, and not Germany? What causes homelessness? Sociologist then devise theories about the answer to their question For example, why the surge in attendance at private academies in Mississippi? Racial competition Legal pressure Mobilization of the civil rights movement Counter-mobilization by white supremacists Each of these theories can be cast into the form of a test able hypothesis Increased competition for a public education from blacks lead to higher PA attendance. Greater legal pressure for desegregation lead to higher PA attendance Greater counter-mobilization by white supremacists leads to higher PA attendance. Notice that each hypothesis has two parts. Y ( X This first part (Y) is called the dependent variable. This variable is considered the effect of the independent variable. The 2nd part (X) is the independent variable. This variable is considered the cause of the something else. Types of variables Continuous variables can take on a large range of values?e.g., income, age. Continuous variables are measured on an interval scale. Categorical variables can take on only a limited number of values--e.g., education Ordinal (e.g., agree, don't know, disagree) Nominal (e.g., religion, race) No order Dummy variables are categorical variables that can take on only two values?e.g., gender. To test a hypothesis, sociologists collect empirical data. This requires them to operationalize their hypothesis. There are two mail types of data that a sociologist can collect Experimental: gold standard, not often feasible Non-experimental: less definitive, but often feasible Experimental method A representative sample is drawn from the population of interest The drug regimen being studied is randomly assigned to individuals in the sample. Random assignment is essential to minimize bias; it permits the attribution of causality to the drug Individuals who are assigned the drug regimen are called the treatment group; individuals who are not randomly assigned a treatment are given a placebo instead; they belong to the control group. Non experimental data Experiments are not possible of rasuble for most sociological research questions Randomly assigning KKK chapters to Mississippi counties comparing the outcomes of treatment and control groups is not possible So sociologists go out into the empirical world and collect as much information as they can Private academy enrollment Racial competition ( proportion of African Americans among school age children Legal pressure ( covered by Alexander? Mobilization --. Number of freedom summer staff Counter mobilization ( KK Analyzing observational data One of the most popular methods for the analysis of this data is regression analysis, which can be used to tell us the ?effect? of one of more independent variables (X) on the dependent variable (Y), all other things being equal. In actuality, multiple regressions is a sophisticated tool for analyzing correlations between independent variable and a dependent variable, but only under very special conditions can it lead to causal inferences. Correlation is no causality Linear regression Linear regression is based on a equation that summarizes the relationship between the independent variables and the dependent variable: y = a + bx Multiple regression Is it possible to have a regression equation with several independent variables? Y = A + B1X1 + B2X2 + B3X3 Andrews? model Academy attendance = + Constant + B1 (proportion black) + B2 (Alexander coverage) + B3 (freedom summer volunteers) + B4 (KKK) What did Andrews find? PA attendance varies positively with the proportion of African Americans among school age children, Alexander coverage number of Freedom Summer volunteers, and the presence of a KKK chapter This lends support to racial competition theory, legal pressure theory, and mobilization/counter mobilization theory Understanding regression results When inspecting a table of regression coefficients look for four things Dependent variable Independent variable Regression coefficients and their signs P-value (i.e., asterisks) Rule of thumb Shows if it is statistically significant