**STAT 330: 98-1
**

Assignment 5 Solutions

- Chapter 6 Q 13: Let be the population mean. Then
so we must evaluate . The definition
is
so that .

- Chapter 6 Q 15: The moment of order
*r*is which is given byMake the substitution for which and get

For the special case

*r*=2 we get the formula in the book (using ).- According to the fact so that . Thus is unbiased.
- You square these numbers, add then divide by 20 and get 74.505 as
in the appendix.

- Chapter 6 Q 28:
- The log of the density is
so that the log likelihood is
To maximize this we take the derivative with respect to and set the result equal to 0 getting

which has root which is the same as the unbiased estimate above.

- The median,
*m*, of the distribution satisfiesThe integral may be done by substituting to get

Solve this to get and then the mle of

*m*is

- The log of the density is
so that the log likelihood is
- Chapter 8 Q 26b: We are given so that the
test rejects when
*t*exceeds the*t*critical value on 7 degrees of freedom. The nearest curve in Table A.13 is for 6 degrees of freedom and the quantity*d*is (4.00-3.5)/1.25= .4 For 6 df this appears to correspond to a of around 0.76 while for 9 df we would get about 0.65. My estimate is about a third of the way between them (because 7 is a third of the way between 6 and 9) or roughly .72. - Chapter 8 Q 30b: See the formula on page 319 and plug in
to get
so that

*n*=24 would be needed. Notice that the formula for the sample size for a two sided level test is just the same as that\ for a one sided level test.It is also acceptable to do this using the t-test graphs as in the previous question. In fact, I think this is what the text intended really. If so you get a sample size of 19.

- Chapter 9 Q 6 b,c: The rejection region of the test is
which can be rewritten as

The power function at is then the area to the right of the right hand side of this last formula under a standard normal curve. Plugging in numbers shows that we want the area to the left of -0.5 which is .31. That is approximately. To get a probability of type II error equal to 0.1 we need

Putting

*m*=40 and plugging all the other numbers we getwhich leads to

*n*of roughly 37 (actually a bit over so rounding up to 38 would be normal). - Chapter 9 Q 12: This is just the formula at the foot of
page 351. We get
which we round up to 50 for safety's sake.

- Chapter 9 Q 73: This is a paired comparison problem. You take
the differences, getting 3.2, -3.4, 0.4, 0.5, 0.3, -1.4, -0.3, 0, 3.5,
-3.7, 3.7, 3.9, 1.6, 3.2. The one sample
*t*statistic is 1.22 on 13 degrees of freedom leading to a 1 sided*P*value of 12% which is only weak evidence of higher TSI for the treatment than for the placebo.

Fri Feb 13 15:37:53 PST 1998