Supplementary Exercise 7.132 of IPS7e
-------------------------------------

Data: Weight loss determined from private weighings prior to and 6 month
after the end of a 10 week program, for 47 subjects.

Analysis: A matched pairs t-test was used, and reported as
"The subjects lost a significant amount of weight over time,
t(46)=4.68, p<0.01".

(a) It is a matched pairs design: two measurements on each subject (before
and after), and analysis of the differences = weight losses.

(b) On the average, the subjects lost weight between the two weighings,
and we are fairly certain that it did not happen by chance alone.

(c) The null hypothesis is H0: mean=0, and the most natural alternative
hypothesis might be Ha: mean>0 (it is of no interest to us whether the
subjects gained weight). In Table C of PSLS we observe that the 99.95% 
percentile of the t(46) distribution must be somewhere between 3.496 and 
3.551 (the values for df=40 and df=30, respectively), so we may
conclude that P<0.0005. From Minitab we get P=0.0000128.


Additional question (critiquing the quote description of the analysis):
-------------------
The main problem with the statement of the results is that it does not
give the average weight loss with either a confidence interval or a
standard error. It would seem as a key information about the weight loss
program how much weight the subjects lost on the average. Statistical
significance does not imply biological significance.


Minitab listing:
----------------
 CDF -4.68;
  T 46.

Student's t distribution with 46 DF

    x  P( X <= x )
-4.68    0.0000128