Homework 2
Due Wednesday, 16 October 1996

  1. For this problem, use the data set estriol.dat (in the Other Data Sets folder).

    Greene and Touchstone (1963) studied the relationship between birthweight and measured hormone levels in pregnant women near the end of their term. The data set contains three variables: a patient identification number, a measurement of estriol production over a 24-hour period (mg/24 hr), and the weight (in 100 gm units) of the baby at birth.

    (a) Construct a scatterplot of birthweight and estriol. (Use graph or plot commands to do this in Stata.)
    (b) Calculate the estimated regression of birthweight on estriol.
    (c) On average, by how much does birthweight increase for each additional mg of estriol produced by the mother?
    (d) What would the predicted birthweight be for a child whose mother had an estriol level of 19 mg/24 hr?
    (e) A colleague asks you to predict what the birthweight would be corresponding to an estriol reading of 35. What would you say?

  2. For this problem, use the data of Selvin, Table 1-4. (This data set can be accessed from the Selvin Data Sets button on the home page for Stat 227.)

    Do boys and girls born to women over 40 differ in birthweight? Use (a) a t-test, (b) an analysis of variance test, and (c) a regression model to address this question. Assuming that bwt and sex refer respectively to the birthweight and infant-sex variables in the data set, the following Stata commands may be helpful:

    ttest bwt, by(sex)
    oneway bwt sex, tabulate
    regress bwt sex

    Comment on the information provided by each of the approaches, and on the extent to which the conclusions drawn from the three approaches are similar to or different from one another.

  3. Selvin, problem 5, page 64.

  4. For this problem, use the data of Selvin, Table 1-4.

    Are there individual variables in the tabulated data that do a better job of predicting birthweight than maternal weight gain (the variable that was discussed in class)? If so, which one(s) and how much better are they? If not, how much better is weight gain than anything else?
Last updated 8-Oct-96
For questions, send mail to r-thisted@uchicago.edu or mcpeek@galton.uchicago.edu