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.

2 . 5<br />

2<br />

1 . 5<br />

1<br />

0 . 5<br />

< 35<br />

35–46 > 46<br />

HDL cholesterol (mg dl –1 )<br />

2.3 Summariz<strong>in</strong>g numerical data 29<br />

< 142<br />

> 198<br />

142–198<br />

Triglyceride<br />

(mg dl –1 )<br />

Fig 2.10 A `three-dimensional' bar diagram show<strong>in</strong>g the relative risk of coronary heart disease <strong>in</strong><br />

men, accord<strong>in</strong>g to high-density lipoprote<strong>in</strong> (HDL) cholesterol and triglyceride (reproduced from<br />

Simons et al., 1991, by permission of the authors and publishers).<br />

to 0 5 cm, still mak<strong>in</strong>g the vertical scale <strong>in</strong> the histogram represent the relative<br />

frequency per cm. We could imag<strong>in</strong>e cont<strong>in</strong>u<strong>in</strong>g this process <strong>in</strong>def<strong>in</strong>itely if there<br />

were no limit to the f<strong>in</strong>eness of the measurement of length or to the number of<br />

observations we could make. In this imag<strong>in</strong>ary situation the histogram would<br />

approach closer and closer to a smooth curve, the frequency curve, which can be<br />

thought of as an idealized form of histogram (Fig. 2.11). The area between the<br />

ord<strong>in</strong>ates erected at any two values of the variable will represent the relative<br />

frequency of observations between these two po<strong>in</strong>ts. These frequency curves are<br />

useful as models on which statistical theory is based, and should be regarded as<br />

idealized approximations to the histograms which might be obta<strong>in</strong>ed <strong>in</strong> practice<br />

with a large number of observations on a variable which can be measured<br />

extremely accurately.<br />

We now consider various features which may characterize frequency distributions.<br />

Any value of the variable at which the frequency curve reaches a peak is<br />

called a mode. Most frequency distributions encountered <strong>in</strong> practice have one<br />

peak and are described as unimodal. For example, the distribution <strong>in</strong> Fig. 2.6 has<br />

a mode at four males, and that <strong>in</strong> Table 2.3 at 40±49 lesions. Usually, as <strong>in</strong> these<br />

two examples, the mode occurs somewhere between the two extremes of the<br />

distribution. These extreme portions, where the frequency becomes low, are<br />

called tails. Some unimodal distributions have the mode at one end of the<br />

range. For <strong>in</strong>stance, if the emission of g-particles by some radioactive material

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

Saved successfully!

Ooh no, something went wrong!