Note: Stata work in this assignment makes use of “gss2012.dta”. Always remember to inspect your data and measures first. When asked to interpret, you do not need to write everything that comes to mind. Instead, try to summarize the findings concisely and meaningfully. Paste all output and syntax.

** This assignment will not be peer graded; however, we will discuss the solutions during lecture and/or TA Stata Labs **

1.   Suppose that you regress children’s reading achievement scores on the number of books in

the household and obtain the following estimates: = 70 and = 0.60.

a.   Interpret the intercept, .

b.   Interpret the slope, .

c.   Re-interpret the slope for an increase of 10 books.

2.   Suppose a colleague sends you a data set that contains two variables, a measure of mental health and a measure of marital satisfaction.  How would you determine which variable should be the dependent variable and which the independent variable?

3.   Suppose that you estimate a multiple regression model using the following variables:

depress             Depression scale (ranging from 0 to 105)

reduc                Respondent's education level (ranging from 0 to 20)

rearninc            Respondent's annual earned income (ranging from $1 to $800,000)

hrwork             Respondent's hours spent at paid work/week (ranging from 1 to 95)

numkid            Number of persons <=18 in the household (ranging from 0 to 10)

And obtain the following results:

a)   Summarize what you learn from these four model fits.

b)  The effect of religion weakens considerably (and even changes direction) after adding

political views to the model. Is it plausible that the relationship between y and religion is spurious? Explain.

9.   Recent UN data from several nations ony = crude birth rate (number of births per 1000 population size), x1  = women’s economic activity (female labor force as percentage of male), and x2 = GNP (per capita, in thousands of dollars) has prediction equation ^(y) =  34.53 — . 13x1  — 0.64x2. The bivariate prediction equation with x1  is  ^(y) = 37.65 — 0.31x1.  The correlations are ryx1   = —0.58 , ryx1    = —0.72, and rx1x2    = 0.58. Explain why the coefficient of x1 in the bivariate equation is quite different from that in the multiple predictor equation.

The following questions use the gss2012_subset file (in Canvas):

10. Perform a multiple linear regression model of respondent’s income in constant dollars (realrinc) on education, mother’s education, and gender. Paste your syntax and output.

Substantively interpret the unstandardized partial coefficients and statistical significance.

11. Now, we'll run the same model that we just performed except that we’ll now treat mother's education as a categorical variable rather than one measured in number of years of education.

Construct a binary variable coded 1 if the respondent's mother received a BA degree or higher and coded as 0 if otherwise. Run the regression and paste your syntax and output. Substantively interpret the partial coefficient for mother's degree (including statistical significance).

12. Now, let's construct a 4-category measure for mother's education; this one categorizes by highest degree. Copy/paste your syntax and output. Report and interpret the partial coefficients for mother's education. What is the reference category?