The Essentials of Biostatistics for Physicians, Nurses, and Clinicians

92 CHAPTER 6 Hypothesis Testing

In this case, we see that the combined p - value is only slightly lower
than the Lynn 2 study.
The fi rst major statistical reference on meta - analysis is Hedges and
Olkin ( 1985 ). Because of the increased popularity of meta - analyses in
medical research, there have been a number of excellent books appear-
ing in recent years. This includes Rothstein et al. ( 2005 ), Hartung et al.
(2008), Whitehead (2002), Stangl and Berry ( 2000 ), and Borenstein
et al. ( 2009 ). Higgins and Green (2008) is a text that provides a summary
of systematic reviews of intervention studies covering numerous meta -
analyses for the Cochrane Group. Michael Borenstein ’ s company has
commercial software to do meta - analysis. Another good recent refer-
ence is Egger et al. (2001).

6.14 EXERCISES

1. DB3 gives baseline serum theophylline levels for patients with emphy-
   sema. Perform an equivalence test to learn if the data are free from a sex
   bias, that is, if mean baseline level is equivalent for men and women.
   There are n 1 = 6 women and n 2 = 10 men. A difference in means of at
   least 2 indicates a bias. The sample means and standard deviations for
   women and men are m 1 = 12.67, s 1 = 3 for the women, and m 2 = 9.68
   and s 2 = 3.65 for the men. You can assume that the standard deviation
   is the same for men and women, and hence use a pooled estimate of the
   standard deviation for the t - test.

Table 6.5

Meta - Analysis of Five Studies of Pig

Blood Loss

Study p - value − 2 ln( p )

Lynn 1 0.44 1.641961

Lynn 2 0.029 7.080919

Martinowitz 0.095 4.714083

Schreiber 1 0.371 1.983106

Schreiber 2 0.086 6.91614

Total 20.33621

Combined 0.026