For the 12-31-90 observation, which day's dummy has a 1? (If more
than one of your day dummies in that row has a 1, you've made an error.)

You are also supposed to put in a dummy variable for the shift in enrollment
which began in mid-year, when Medicaid eligibility was expanded.
What is the value of your shift dummy for the 12-31-90 observation?

If you got all those right, there is a good chance that your dummy variables are OK.

Let's spot-check your regression results:

What did you get for the coefficient of Time? (The first four digits to the right of the decimal point have to be right.)

Based on this coefficient, its standard error, the T-statistic (which is
the coefficient divided by the standard error), and the P-value associated
with that T-statistic, pick the best way to complete this sentence:

If there had been no eligibility change at mid-year, the daily number of
Medicaid-paid births during the year 1990 would have ...

When you did the regression the first time, you had to leave out one of
these to avoid multicolinearity among your independent variables.

Now for the prediction for 1991. You're using your spreadsheet to make the predictions, rather than LS, because you need a prediction for every day of 1991. That's 365 predictions! A spreadsheet is better than LS for this.

In your spreadsheet, extend your dummy variable columns down another 365 rows. Also, extend the Time variable down 365 rows.

Then use the estimated coefficients from your regression to create a spreadsheet formula for a predicted value. The formula in cell B367 will be of the form

=123+123*C367+123*D367+123*E367+123*F367+123*G367+123*H367+123*I367+123*J367+123*K367

If you copy and paste that formula, it will give you wrong answers unless you edit it.

- The 123s are placeholders for numbers that you get from your regression results.
- Your estimated intercept goes where the first 123 is.
- Your other estimated coefficients go where the other 123's are.
- If you left Monday out of your equation, leave out the E367 term.
- If you left Tuesday out of your equation, leave out the F367 term.
- If you left Wednesday out of your equation, leave out the G367 term.
- If you left Thurseday out of your equation, leave out the H367 term.
- If you left Friday out of your equation, leave out the I367 term.
- If you left Saturday out of your equation, leave out the J367 term.
- If you left Sunday out of your equation, leave out the K367 term.

That formula, in B367, gives you your predicted births for 1-1-91.

What did you get for predicted births for 1-1-91?

If you got that one right, you can copy the formula in B367 all the way down the column to B731. This gives you 365 predictions, one for each day of 1991.

Add up the 1991 daily predictions, and you get the total for the year:

Now you can write your memo. Remember, your boss knows more than
the pointy-haired guy in Dilbert(TM), but not much more.

That's all, folks!

Copyright © 1999-2003 Samuel L. Baker

The views and opinions expressed in this page are strictly those of the page author. The contents of this page have not been reviewed or approved by the University of South Carolina.

Please e-mail comments to