Download PDF
Methods Inf Med 1988; 27(01): 37-42
DOI: 10.1055/s-0038-1635514
Original Article
Schattauer GmbH

Influence of Sampling Variation and Analytical Errors on the Performance of the Multivariate Reference Region

Der Einfluß von Stichprobenstreuung und Analysefehlern auf die Leistung des multivariaten Referenzbereichs
K. Linnet
1   (From the Department of Clinical Chemistry, State University Hospital, Rigshospitalet, Copenhagen, Denmark)
› Author Affiliations
Further Information

Publication History

Publication Date:
20 February 2018 (online)

  PDF Download  Permissions and Reprints

Summary

The effects of sampling variation, of deviations from normality of data distributions, and of analytical errors on the performance of the parametric multivariate 0.95 reference region are studied. The standard error of the proportion being included in the region is similar to that of the univariate interval, i.e. 0.016 for N = 100. Deviations from normality cause a moderate decrease of the proportion included in the region.

The region is very sensitive to an increase of the analytical imprecision occurring after the region has been established. E.g., for a dimension of 20, an increase of the total standard deviation of each analyte distribution by a factor of 1.25 decreases the included proportion from 0.95 to 0.54. Analytical inaccuracy may have a strong impact when the variables are correlated. In conclusion, stable analytical methods and efficient quality control systems are important to ensure a reasonable performance of a multivariate reference region.

Die Auswirkungen der Stichprobenstreuung, der Abweichung von der Normalverteilung und der Analysefehler auf die Leistung multivariater parametrischer 95% Referenzbereiche werden untersucht. Der Standardfehler des in den Bereich fallenden Anteils ist ähnlich dem des univariaten Intervalls, d.h. 0,016 für N = 100. Abweichungen von der Normalverteilung verursachen eine geringfügige Abnahme des in den Bereich fallenden Anteils.

Der Bereich reagiert sehr empfindlich auf eine Zunahme der analytischen Ungenauigkeit, die nach Festlegung des Bereichs eintritt. So vermindert z.B. für einen Umfang von 20 eine Zunahme der Gesamtstandardabweichung jeder Analyseverteilung um einen Faktor 1,25 den einbezogenen Anteil von 0,95 auf 0,54. Analytische Ungenauigkeit kann bei korrelierten Variablen einen starken Einfluß ausüben.

Abschließend ist zu sagen, daß stabile analytische Methoden und wirksame Systeme der Qualitätskontrolle für die Sicherstellung einer annehmbaren Leistung eines multivariaten Referenzbereichs wichtig sind.

Key-Words:

Analytical Imprecision - Inaccuracy - Multivariate Reference Region - Non-normality - Quality Control

Schlüssel-Wörter:

Analytische Ungenauigkeit - multivariater Referenzbereich - Nichtnormali-tät - Qualitätskontrolle - Standardfehler - Statistik
 

  • REFERENCES

  • 1 Aronsson T, Bjørnstad P, Leskinen E, Uldall A, de Verdier C-H. Assessment of analytical quality in Nordic clinical chemistry laboratories using data from contemporary national programs. Scand. J. Clin. Lab. Invest 1984; 172 (Suppl. 44) Suppl 115-124.
    PubMedGoogle Scholar
  • 2 Box G. E. P, Cox D. R. An analysis of transformations. J. roy. statist. Soc, Ser. B 1964; 26: 211-252.
    PubMedGoogle Scholar
  • 3 Boyd J. C, Lacher D. A. The multivariate reference range: An alternative interpretation of multitest profiles. Clin. Chem 1982; 28: 259-265.
    PubMedGoogle Scholar
  • 4 Chew V. Confidence, prediction, and tolerance regions for the multivariate normal distribution. J. amer. stat. Ass 1966; 61: 605-617.
    CrossrefPubMedGoogle Scholar
  • 5 Davies T. C, Griffin K. E. Use of Mahalanobis generalized distance in interpreting low-resolution gas-chromatographic profiles of urinary steroids. Clin. Chem 1986; 32: 2016-2020.
    PubMedGoogle Scholar
  • 6 Durbridge T. C. Clinical acceptance of a multi-test reference region for biochemicalpanel results. Clin. Chem 1983; 29: 1724-1726.
    PubMedGoogle Scholar
  • 7 El-Sheikh T. S, Wacker A. G. Effect of dimensionality and estimation on the performance of Gaussian classifiers. Pattern Recogn 1980; 12: 115-126.
    CrossrefPubMedGoogle Scholar
  • 8 Fraser C. G. Analytical goals in clinical biochemistry. Progr. clin. Pathol 1981; 8: 101-122.
    PubMedGoogle Scholar
  • 9 Grams R. R, Johnson E. A, Benson E. S. Laboratory data analysis system: Section III – multivariate normality. Amer. J. clin. Pathol 1972; 58: 188-200.
    PubMedGoogle Scholar
  • 10 Gräsbeck R, Alström T. (Eds) Reference Values in Laboratory Medicine. New York: Wiley; 1981
    Google Scholar
  • 11 Harris E. K. Statistical aspects of reference values in clinical pathology. Progr. clin. Pathol 1981; 8: 45-66.
    PubMedGoogle Scholar
  • 12 Johnson N. L, Pearson E. S. Tables of percentage points of non-central X . Biometrika 1969; 56: 255-272.
    CrossrefPubMedGoogle Scholar
  • 13 Kågedal B, Sandström A, Tibbling G. Determination of a trivariate reference region for free thyroxine index, free triiodothyronine index, and thyrotropin from results obtained in a health survey of middleaged women. Clin. Chem 1978; 24: 1744-1750.
    PubMedGoogle Scholar
  • 14 Linnet K. Two-stage transformation systems for normalization of reference distributions evaluated. Clin. Chem 1987; 33: 381-386.
    PubMedGoogle Scholar
  • 15 Numerical Algorithms Group (NAG): Fortran Library Manual. Numerical Algorithms Group Ltd. NAG Central Office, Mayfield House, 256 Banbury Road, Oxford OX2 7DE, UK
    PubMed
  • 16 Rehpenning W, Harm K, Domeste A, Voigt K. D. Falsch positive Werte bei der Vielfachanalyse: Die Abschätzung ihrer Häufigkeit mit der Sylvesterschen Formel und ihre Reduktion durch eine multivariate Testgröße. J. clin. Chem. clin. Biochem 1979; 17: 565-572.
    PubMedGoogle Scholar
  • 17 Scheuer E. M, Stoller D. S. On the generation of normal random vectors. Technometrics 1962; 4: 278-281.
    CrossrefPubMedGoogle Scholar
  • 18 Solberg H. E. Discriminant analysis. Crit. Rev. clin. Lab. Sei 1978; 9: 209-242.
    PubMedGoogle Scholar
  • 19 Westgaard J. O, Groth T. Power functions for statistical control rules. Clin. Chem 1979; 25: 863-869.
    PubMedGoogle Scholar
  • 20 Westgaard J. O, Barry P. L, Hunt M. R, Groth T. A multirule Shewhart chart for quality control in clinical chemistry. Clin. Chem 1981; 27: 493-501.
    PubMedGoogle Scholar
  • 21 Wilks S. S. Determination of sample sizes for setting tolerance limits. Ann. Math. Statist 1941; 12: 91-96.
    CrossrefPubMedGoogle Scholar
  • 22 Winkel P, Lyngbye J, Jörgensen K. The normal region – a multivariate problem. Scand. J. clin. Lab. Invest 1972; 30: 339-344.
    CrossrefPubMedGoogle Scholar
  • 23 Winkel P. Patterns and clusters – multivariate approach for interpreting clinical chemistry results. Clin. Chem 1973; 19: 1329-1338.
    PubMedGoogle Scholar