01.06.2013 Views

Statistical Methods in Medical Research 4ed

Statistical Methods in Medical Research 4ed

Statistical Methods in Medical Research 4ed

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

Equations (11.21) and (11.22) can easily be seen to be equivalent to (11.16) and<br />

(11.17) if, <strong>in</strong> the calculation of wi <strong>in</strong> (11.14), the separate estimates of residual<br />

variance s 2 i are replaced by the common estimate s2 . For tests or the calculation<br />

of confidence limits for b us<strong>in</strong>g (11.22), the t distribution on n1 ‡ n2 4DF<br />

should be used. Where a common slope is accepted it would be more usual to<br />

estimate s 2 as the residual mean square about the parallel l<strong>in</strong>es (11.34), which<br />

would have n1 ‡ n2 3 DF.<br />

Example 11.3<br />

Table 11.4 gives age and vital capacity (litres) for each of 84 men work<strong>in</strong>g <strong>in</strong> the cadmium<br />

<strong>in</strong>dustry. They are divided <strong>in</strong>to three groups: A1, exposed to cadmium fumes for at least<br />

10 years; A2, exposed to fumes for less than 10 years; B, not exposed to fumes. The ma<strong>in</strong><br />

purpose of the study was to see whether exposure to fumes was associated with a change <strong>in</strong><br />

respiratory function. However, those <strong>in</strong> group A1 must be expected to be older on the<br />

average than those <strong>in</strong> groups A2 or B, and it is well known that respiratory test performance<br />

decl<strong>in</strong>es with age. A comparison is therefore needed which corrects for discrepancies<br />

between the mean ages of the different groups.<br />

We shall first illustrate the calculations for two groups by amalgamat<strong>in</strong>g groups A1<br />

and A2 (denot<strong>in</strong>g the pooled group by A) and compar<strong>in</strong>g groups A and B.<br />

The sums of squares and products of deviations about the mean, and the separate<br />

slopes bi are as follows:<br />

Group i ni Sxxi Sxyi Syyi bi<br />

A 1 40 4397 38 236 385 26 5812 0 0538<br />

B 2 44 6197 16 189 712 20 6067 0 0306<br />

Total 10594 54 426 097 47 1879 ( 0 0402)<br />

The SSq about the regressions are<br />

P<br />

…1† …y Y1† 2 ˆ 26 5812 … 236 385† 2 =4397 38 ˆ 13 8741<br />

and<br />

P<br />

…2† …y Y2† 2 ˆ 20 6067 … 189 712† 2 =6197 16 ˆ 14 7991:<br />

Thus,<br />

s 2 ˆ…13 8741 ‡ 14 7991†=…40 ‡ 44 4† ˆ0 3584,<br />

and, for the difference between b1 and b2 us<strong>in</strong>g (11.19) and (11.20),<br />

0 0538 … 0 0306†<br />

t ˆ<br />

1<br />

…0 3584†<br />

4397 38 ‡<br />

r<br />

1<br />

6197 16<br />

ˆ 0 0232=0 0118<br />

ˆ 1 97 on 80 DF:<br />

11.4 Regression <strong>in</strong> groups 325

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!