25.11.2014 Views

Biostatistics

Create successful ePaper yourself

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

728 CHAPTER 13 NONPARAMETRIC AND DISTRIBUTION-FREE STATISTICS<br />

Theil’s Slope Estimator Theil (12) proposes a method for obtaining a point<br />

estimate of the slope coefficient b. We assume that the data conform to the classic<br />

regression model<br />

y i ¼ b 0 þ b 1 x 1 þ e i ;<br />

i ¼ 1; ...; n<br />

where the x i are known constants, b 0 and b 1 are unknown parameters, and Y i is an observed<br />

value of the continuous random variable Y at x i . For each value of x i , we assume a<br />

subpopulation of Y values, and the e i are mutually independent. The x i are all distinct (no<br />

ties), and we take x 1 < x 2 < < x n .<br />

The data consist of n pairs of sample observations, ðx 1 ; y 1 Þ; ðx 2 ; y 2 Þ; ...; ðx n ; y n Þ,<br />

where the ith pair represents measurements taken on the ith unit of association.<br />

To obtain Theil’s estimator of b 1 we first form all possible sample slopes<br />

S ij ¼ y j y i = xj x i , where i < j. There will be N ¼ n C 2 values of S ij . The estimator<br />

of b 1 which we designate by ^b 1 is the median of S ij values. That is,<br />

<br />

^b 1 ¼ median S ij<br />

The following example illustrates the calculation of ^b 1 .<br />

(13.11.1)<br />

EXAMPLE 13.11.1<br />

In Table 13.11.1 are the plasma testosterone (ng/ml) levels (Y) and seminal citric acid<br />

(mg/ml) levels in a sample of eight adult males. We wish to compute the estimate of the<br />

population regression slope coefficient by Theil’s method.<br />

Solution: The N ¼ 8 C 2 ¼ 28 ordered values of S ij are shown in Table 13.11.2.<br />

If we let i ¼ 1 and j ¼ 2, the indicators of the first and second values of<br />

Y and X in Table 13.11.1, we may compute S 12 as follows:<br />

S 12 ¼ ð175 230Þ= ð278 421Þ ¼ :3846<br />

When all the slopes are computed in a similar manner and ordered as<br />

in Table 13.11.2, :3846 winds up as the tenth value in the ordered<br />

array.<br />

The median of the S ij values is .4878. Consequently, our estimate of the<br />

population slope coefficient ^b 1 ¼ :4878.<br />

TABLE 13.11.1 Plasma Testosterone and Seminal Citric Acid<br />

Levels in Adult Males<br />

Testosterone: 230 175 315 290 275 150 360 425<br />

Citric acid: 421 278 618 482 465 105 550 750

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

Saved successfully!

Ooh no, something went wrong!