26.12.2014 Views

Clinical Biochemistry of Domestic Animals (Sixth Edition) - UMK ...

Clinical Biochemistry of Domestic Animals (Sixth Edition) - UMK ...

Clinical Biochemistry of Domestic Animals (Sixth Edition) - UMK ...

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

6<br />

Chapter | 1 Concepts <strong>of</strong> Normality in <strong>Clinical</strong> <strong>Biochemistry</strong><br />

Anderson-Darling normality test<br />

A-squared 3.39<br />

P-value 0.005<br />

Mean 66.042<br />

StDev 31.447<br />

Variance 988.902<br />

Skewness 0.929303<br />

Kurtosis 0.549188<br />

N 168<br />

30<br />

60<br />

90<br />

120<br />

150<br />

180<br />

Minimum 18.000<br />

1st quartile 43.000<br />

Median 58.500<br />

3rd quartile 83.750<br />

Maximum 184.000<br />

95% Confidence intervals<br />

* *<br />

95% Confidence interval for mean<br />

61.252 70.832<br />

95% Confidence interval for median<br />

53.000 64.000<br />

95% Confidence interval for StDev<br />

28.406 35.223<br />

Mean<br />

Median<br />

55<br />

60<br />

65<br />

70<br />

FIGURE 1-3 Distribution and summary statistics for the sample <strong>of</strong> canine alanine aminotransferase values<br />

(U/liter) in Table 1-1 . Printout <strong>of</strong> MINITAB, Release 14.13.<br />

1 . Use <strong>of</strong> Transformations<br />

Frequently, some transformation (such as the logarithmic or<br />

square root transformation) <strong>of</strong> the analyte values will make<br />

the distribution more Gaussian ( Kleinbaum et al. , 2008 ;<br />

Neter et al. , 1996 ; Zar, 1999 ). The boundaries for the reference<br />

values are two standard deviations above and below<br />

the mean for the distribution <strong>of</strong> the transformed analyte<br />

values. These boundaries then can be expressed in terms <strong>of</strong><br />

the original analyte values by retransformation. Figure 1-4<br />

describes the distribution <strong>of</strong> the ALT analyte values after<br />

transformation with natural logarithms. The reference<br />

boundaries in logarithmic units are equal to 4.08013 <br />

(1.96 0.47591) or (3.14734, 5.01292), which correspond<br />

to (23.3, 150.3U/liter), in the original units <strong>of</strong> the analyte.<br />

2 . Use <strong>of</strong> Percentiles<br />

The second approach that can be followed in the situation<br />

where an assumption <strong>of</strong> a Gaussian distribution is not tenable<br />

is to choose percentiles as boundaries ( Feinstein, 1977 ;<br />

Herrera, 1958 ; Mainland, 1963 ; Massod, 1977 ; Reed et al. ,<br />

1971 ; Solberg, 1999 ). For example, if we wanted to misclassify<br />

only 5% <strong>of</strong> normal animals as being abnormal, the<br />

2.5th and 97.5th percentiles could be chosen as the reference<br />

boundaries. Thus, animals would be classified as abnormal<br />

when having analyte values either below the value <strong>of</strong> the<br />

analyte below which are 2.5% <strong>of</strong> all normal analyte values<br />

or above the value <strong>of</strong> the analyte below which are 97.5% <strong>of</strong><br />

all normal analyte values. This method is attractive because<br />

percentiles are reflective <strong>of</strong> the distribution involved.<br />

The 97.5th percentile is estimated as the value <strong>of</strong> the<br />

analyte corresponding to the ( n 1) 0.975th observation<br />

in an ascending array <strong>of</strong> the analyte values for a sample <strong>of</strong> n<br />

normal animals ( Dunn and Clark, 2001 ; Ryan et al. , 2001 ;<br />

Snedecor and Cochran, 1989 ). For the ALT values from<br />

the sample <strong>of</strong> n 168 animals, (n 1) 0.975 169 <br />

0.975 164.775. Because there is no 164.775th observation,<br />

the 97.5th percentile is found by interpolating between<br />

the ALT values corresponding to the 164th and 165th<br />

observation in the ascending array commonly referred<br />

to as the 164th and 165th order statistics ( Ryan et al .,<br />

2001 ; Snedecor and Cochran, 1989 ). The 164th order statistic<br />

is 138U/liter and the 165th order statistic is 140U/<br />

liter and the interpolation is 138 0.775(140 138) <br />

139.5U/liter. The 2.5th percentile is estimated similarly as<br />

the ( n 1) 0.025th order statistic, which is the 4.225th<br />

order statistic for the sample <strong>of</strong> ALT values. In this case,<br />

the 4th and 5th order statistics are the same, 24 U/liter,<br />

which is the estimate <strong>of</strong> the 2.5th percentile. Note that there<br />

is reasonable agreement between this reference interval and

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

Saved successfully!

Ooh no, something went wrong!