Sociology 709

PS10

Note: As part of this homework, I am going to ask you to try the “ice” command in Stata. If you find this command too hard to use, just fall back to the impute command. Example syntax for using both commands is in the example do file for the lecture on missing data.

If you have your own copy of Stata, you can install ice by typing “findit ice” and following the links, or by simply typing:

net sj 5-4 st0067_2

in Stata.

If you are working on a public computer (or you are using Stata over a network), you will not have write access to system directories, but you can install ice to your own directory. Follow these directions. Again, if you find this process too difficult just use impute.

1. Using the data set ps10_nlsy1 estimate the following regression equation:

xi: reg lnwage hgc i.sex age age2 exp80 exp802 ten*

using the impute command or ice to deal with the missing data. [Note that the true result of this regression is the actual data column of the table from the lecture].

Steps: a. type “sum” to see what variables have missing data.

b. If you are using impute, follow the example from the do-file for lecture to impute the data. The run the regression (see the syntax from lecture).

c. If you are using ice, follow the example from lecture to run the multiple imputation, then use the micombine command to estimate the regression (again, see the syntax from lecture).

2. Briefly, explain why multiple imputation is better than single imputation.

3. Longitudinal data: Write and run a dofile with the following steps:

a. In Stata, type “set mem 30m” to give yourself enough memory.

b. use the data set nlsy_seg.dta

c. run the random effects model, xi: xtreg lnwage i.sex pctfem hgc hrswk tenur tenfem exp* , re

compare your results to the results from the fixed effects model (from lecture).

d. Does the effect of job tenure (tenur) on log wages differ between men and women? Test this using a fixed effects approach.

4. Briefly, explain the chief advantage of a fixed effects model, as discussed in lecture. Describe the real or hypothetical use of such a model in a research field you are interested in.