Expand Your Knowledge: Software Approximation for Degrees of Freedom Given x1 and x2 distributions that are normal or approximately normal with unknown s1 and s2, the value of t corresponding to x1 2x2 has a distribution that is approximated by a Student’s t distribution. We use the convention that the degrees of freedom are approximately the smaller of n1 21 and n2 21. However, a more accurate estimate for the appropriate degrees of freedom is given by Satterthwaite’s formula:

where, s1, s2, n1 and n2 are the respective sample standard deviations and sample sizes of independent random samples from the x1 and x2 distributions. This is the approximation used by most statistical software.
When both n1 and n2 are 5 or larger, it is quite accurate.
The degrees of freedom computed from this formula are either truncated or not rounded.

(a) In Problem 21, we tested whether the population average crime rate m2 in the Rocky Mountain region is higher than that in New England, m1. The data were n1 510, x1

(b) When you did Problem 21, you followed the convention that degrees of freedom

d. f . 5smaller of n1 51 and n2 51. Compare this value of d.f. with that found with Satterthwaite’s formula.

AppendixLO1

2 d.f. 225 1 S n-1 1 n n 2 + + $ 12 1 - 2