Abstract
We provide a fundamental theorem that can be used in conjunction with Kolmogorov asymptotic conditions to derive the first moments of well-known estimators of the actual error rate in linear discriminant analysis of a multivariate Gaussian model under the assumption of a common known covariance matrix. The estimators studied in this paper are plug-in and smoothed resubstitution error estimators, both of which have not been studied before under Kolmogorov asymptotic conditions. As a result of this work, we present an optimal smoothing parameter that makes the smoothed resubstitution an unbiased estimator of the true error. For the sake of completeness, we further show how to utilize the presented fundamental theorem to achieve several previously reported results, namely the first moment of the resubstitution estimator and the actual error rate. We provide numerical examples to show the accuracy of the succeeding finite sample approximations in situations where the number of dimensions is comparable or even larger than the sample size.
Similar content being viewed by others
References
Anderson, T. (1951). Classification by multivariate analysis. Psychometrika, 16, 31–50.
Anderson, T. (1973). An asymptotic expansion of the distribution of the studentized classification statistic w. Ann. Statist., 1, 964–972.
Conte, E., Lops, M., and Ricci, G. (1996). Adaptive matched filter detection in spherically invariant noise. IEEE Signal Process. Lett., 3, 248–250.
Deev, A. (1970). Representation of statistics of discriminant analysis and asymptotic expansion when space dimensions are comparable with sample size. Dokl. Akad. Nauk SSSR, 195, 759–762 (in Russian).
Deev, A. (1972). Asymptotic expansions for distributions of statistics w, m, and w* in discriminant analysis. Statist. Methods Class., 31, 6–57 (in Russian).
Dougherty, E.R. (2008). On the epistemological crisis in genomics. Curr. Genomics, 9, 69–79.
Dougherty, E.R., Hua, J., and Bittner, M. (2007). Validation of computational methods in genomics. Curr. Genomics, 8, 1–19.
Dunn, O.J. (1971). Some expected values for probabilities of correct classification in discriminant analysis. Technometrics, 13, 345–353.
Dupuy, A. and Simon, R. (2008). Critical review of published microarray studies for cancer outcome and guidelines on statistical analysis and reporting. J. Natl. Cancer Inst., 99, 147–157.
Efron, B. (1975). The efficiency of logistic regression compared to normal discriminant analysis. J. Amer. Statist. Assoc., 70, 892–898.
Efron, B. (1980). The distributions of the actual error rates in linear discriminant analysis. J. Amer. Statist. Assoc., 75, 201–205.
Fisher, R. (1936). The use of multiple measurements in taxonomic problems. Ann. Euge., 7, 79–188.
Fisher, R. (1940). The precision of discriminant function. Ann. Euge., 10, 422–429.
Fisher, R.A. (1925). Statistical methods for research workers, 14th edn. Oliver & Boyd, Edinburgh. The quotation is from the preface to the first (1925) edition.
Foley, D. (1972). Considerations of sample and feature size. IEEE Trans. Inform. Theory, IT-18, 618–626.
Fujikoshi, Y. (2000). Error bounds for asymptotic approximations of the linear discriminant function when the sample sizes and dimensionality are large. J. Multivariate Anal., 73, 1–17.
Fujikoshi, Y. and Seo, T. (1998). Asymptotic approximations for epmc’s of the linear and the quadratic discriminant functions when the samples sizes and the dimension are large. Statist. Anal. Random Arrays, 6, 269–280.
Fukunaga, K. and Hayes, R.R. (1989). Estimation of classifier performance. IEEE Trans. Pattern Anal. Mach. Intell., 11, 1087–1101.
Gevaert, O., Smet, F.D., Gorp, T.V., Pochet, N., Engelen, K., Amant, F., Moor, B.D., Timmerman, D. and Vergote, I. (2008). Expression profiling to predict the clinical behaviour of ovarian cancer fails independent evaluation. BMC Cancer, 8, 1–10.
Glick, N. (1978). Additive estimators for probabilities of correct classification. Pattern Recognit., 10, 211–222.
Hand, D. (1986). Recent advances in error rate estimation. Pattern Recognit. Lett., 4, 335–346.
Hills, M. (1966). Allocation rules and their error rates. J. R. Stat. Soc. Ser. B (Methodological), 28, 1–31.
Hirst, D. (1996). Error-rate estimation in multiple-group linear discriminant analysis. Technometrics, 38, 389–399.
John, S. (1961). Errors in discrimination. Ann. Math. Stat., 32, 1125–1144.
Johnson, N., Kotz, S. and Balakrishnan, N. (1994). Continuous univariate distributions. John Wiley, New York.
Kim, S., Dougherty, E.R., Shmulevich, I., Hess, K.R., Hamilton, S.R., Trent, J.M., Fuller, G.N. and Zhang, W. (2002). Identification of combination gene sets for glioma classification. Mol. Cancer Ther., 1, 1229–1236.
Kittler, J. and Devijver, P. (1982). Statistical properties of error estimators in performance assessment of recognition systems. IEEE Trans. Pattern Anal. Mach. Intell., 4, 215–220.
Lachenbruch, P. and Mickey, M. (1968). Estimation of error rates in discriminant analysis. Technometrics, 10, 1–11.
Martin, J.K. and Hirschberg, D.S. (1996). Small Sample Statistics for Classification Error Rates II: Confidence Intervals and Significance Tests. Tech. Rep. 96-22, University of California, Irvine, CA.
Mclachlan, G.J. (1973). An asymptotic expansion of the expectation of the estimated error rate in discriminant analysis. Aust. J. Statistics, 15, 210–214.
Mclachlan, G.J. (1974). Estimation of the errors of misclassification on the criterion of asymptotic mean square error. Technometrics, 16, 255–260.
Mclachlan, G.J. (1976). The bias of the apparent error in discriminant analysis. Biometrika, 63, 239–244.
Mclachlan, G.J. (1992). Discriminant analysis and statistical pattern recognition. John Wiley, New York.
Meshalkin, L.D. and Serdobolskii, V.I. (1978). Errors in the classification of multi-variate observations. Theory Probab. Appl., 23, 741–750.
Michiels, C.H.S. and Koscielny, S. (2005). Prediction of cancer outcome with microarrays: a multiple random validation strategy. Lancet, 365, 488–492.
Moran, M. (1975). On the expectation of errors of allocation associated with a linear discriminant function. Biometrika, 62, 141–148.
Moreira, M. (2009). Tests with correct size when instruments can be arbitrarily weak. J. Econometrics, 152, 131–140.
Okamoto, M. (1963). An asymptotic expansion for the distribution of the linear discriminant function. Ann. Math. Stat., 34, 1286–1301 (Correction: Ann. Math. Stat., 39, 1358–1359, 1968).
Raudys, S. (1967). On determining training sample size of a linear classifier. Comput. Syst., 28, 79–87 (in Russian).
Raudys, S. (1972). On the amount of a priori information in designing the classification algorithm. Tech. Cybern., 4, 168–174 (in Russian).
Raudys, S. (1978). Comparison of the Estimates of the Probability of Misclassification. In Proc. International Joint Conference on Pattern Recognition, pp 280–282.
Raudys, S. (1998). Expected classification error of the fisher linear classifier with pseudo-inverse covariance matrix. Pattern Recognit. Lett., 19, 385–392.
Raudys, S. and Jain, A.K. (1991). Small sample size effects in statistical pattern recognition: recommendations for practitioners. IEEE Trans. Pattern Anal. Mach. Intell., 13, 252–264.
Raudys, S. and Pikelis, V. (1980). On dimensionality, sample size, classification error, and complexity of classification algorithm in pattern recognition. IEEE Trans. Pattern Anal. Mach. Intell., 2, 242–252.
Raudys, S. and Skurikhina, M. (1995). Small-sample properties of ridge-estimate of the covariance matrix in statistical and neural net classification. Multivariate Statist., 237–245.
Schiavo, R.A. and Hand, D.J. (2000). Ten more years of error rate research. Internat. Statist. Rev., 68, 295–310.
Serdobolskii, V. (2000). Multivariate statistical analysis: a high-dimensional approach. Kluwer Academic Publishers, Netherlands.
Serdobolskii, V.I. (1979). The Moments of Discriminant Function and Classification for a Large Number of Variables. In Statistical Problems of Control (S. Raudys, ed.). Inst. of Math. and Cyb. Press, Vilnius, pp 27–51, in Russian.
Smith, C. (1947). Some examples of discrimination. Ann. Euge., 18, 272–282.
Snapinn, S. and Knoke, J. (1985). An evaluation of smoothed classification error-rate estimators. Technometrics, 27, 199–206.
Snapinn, S. and Knoke, J. (1989). Estimation of error rates in discriminant analysis with selection of variables. Biometrics, 45, 289–299.
Sorum, M.J. (1971). Estimating the conditional probability of misclassification. Technometrics, 13, 333–343.
Sorum, M.J. (1972). Estimating the expected and the optimal probabilities of misclassification. Technometrics, 14, 935–943.
Sorum, M.J. (1973). Estimating the expected probability of misclassification for a rule based on the linear discriminant function: Univariate normal case. Technometrics, 15, 329–339.
Swets, D. and Weng, J. (1996). Using discriminant eigenfeatures for image retrieval. IEEE Trans. Pattern Anal. Mach. Intell., 18, 891–896.
Van Vuuren, S. and Hermansky, H. (1997). Data-driven design of rasta-like filters. In: Proc. Eurospeech, 1607–1610.
Wacker, A. and El-Sheikh, T. (1984). Average classification accuracy over collections of Gaussian problems—common covariance matrix case. Pattern Recognit., 17, 259–274.
Wald, A. (1944). On a statistical problem arising in the classification of an individual into one of two groups. Ann. Math. Stat., 15, 145–162.
Wigner, E.P. (1958). On the distribution of the roots of certain symmetric matrices. Ann. Math., 67, 325–327.
Zollanvari, A., Braga-Neto, U. and Dougherty, E. (2009). On the sampling distribution of resubstitution and leave-one-out error estimators for linear classifiers. Pattern Recognit., 42, 2705–2723.
Zollanvari, A., Braga-Neto, U. and Dougherty, E. (2011). Analytic study of performance of error estimators for linear discriminant analysis. IEEE Trans. Signal Process., 59, 4238–4255.
Author information
Authors and Affiliations
Corresponding author
Rights and permissions
About this article
Cite this article
Zollanvari, A., Genton, M.G. On Kolmogorov asymptotics of estimators of the misclassification error rate in linear discriminant analysis. Sankhya A 75, 300–326 (2013). https://doi.org/10.1007/s13171-013-0029-9
Received:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s13171-013-0029-9
Keywords and phrases.
- Double asymptotics
- error estimation
- Kolmogorov asymptotic analysis
- plug-in error
- resubstitution
- smoothed resubstitution
- true error