Stepwise Variable Selection in Classification Problems

Two variable selection procedures are evaluated for classification problems: a forward stepwise discrimination procedure, and a stepwise procedure preceded by a preliminary screening of variables on the basis of individual t statistics. Expected probability of correct classification is used as the measure of performance. A comparison is made of the procedures using samples from multi-variate normal populations and from several nonnormal populations. The study demonstrated some situations where the use of all variables is preferable to the use of a stepwise discriminant procedure stopping after a few steps, though usually the latter procedure was superior in performance. However where the stepwise procedure performed better than using all variables, the modified stepwise procedure performed still better. The use of modified stepwise procedures in which not all the covariances of the problem need be estimated seems promising.     

A thorough and quantified method for classifying seabird diving behaviour

Time-depth recorders are commonly deployed on diving animals to obtain information on their aquatic behaviour. The recorded data provide a 2D profile of diving activity. As analyses of diving behaviour from such profiles have become more complex, these analyses have often suffered from a lack of consistency and rigour. There is a growing need for a simple, comparative method to classify diving behaviour thoroughly and quantitatively. Here, a new approach to the classification of the dive profiles of penguins is described, which probably has applicability for many other diving predators as well. This simple approach uses a small, coherent set of criteria to classify behaviours in a detailed and quantified manner, and with relative objectivity. Classification of diving behaviour is possible from the temporal scale of a wiggle within a dive to the scale of a bout of dives. The new method will make comparisons between species easier and clearer because these comparisons will be undertaken within a consistent, more objective framework.     

A Two-Way Analysis of Covariance Model for Classification of Stability Data

This paper proposes a procedure for testing and classifying data with multiple factors. A two-way analysis of covariance is used to classify the differences among the batches as well as another factor such as package type and/or product strength. In the test procedure, slopes and intercepts of the main effects are tested using a combination of simultaneous and sequential F-tests. Based on the test procedure results, the data are classified into one of four different groups. For each group, shelf life can be calculated accordingly. We examine if the procedure produces satisfactory control of the probability of a Type I error and the power of detecting the difference of degradation rates and intercepts for different nominal levels. The method is evaluated with a Monte Carlo simulation study. The proposed procedure is compared with the current FDA procedure using real data.     

Analysis of donor splice sites in different eukaryotic organisms

We present here a new algorithm for functional site analysis. It is based on four main assumptions: each variation of nucleotide composition makes a different contribution to the overall binding free energy of interaction between a functional site and another molecule; nonfunctioning site-like regions (pseudosites) are absent or rare in genomes; there may be errors in the sample of sites; and nucleotides of different site positions are considered to be mutually dependent. In this algorithm, the site set is divided into subsets, each described by a certain consensus. Donor splice sites of the human protein-coding genes were analyzed. Comparing the results with other methods of donor splice site prediction has demonstrated a more accurate prediction of consensus sequences AG/GU(A,G), G/GUnAG, /GU(A,G)AG, /GU(A,G)nGU, and G/GUA than is achieved by weight matrix and consensus (A,C)AG/GU(A,G)AGU with mismatches. The probability of the first type error, E1, for the obtained consensus set was about 0.05, and the probability of the second type error, E2, was 0.15. The analysis demonstrated that accuracy of the functional site prediction could be improved if one takes into account correlations between the site positions. The accuracy of prediction by using human consensus sequences was tested on sequences from different organisms. Some differences in consensus sequences for the plant Arabidopsis sp., the invertebrate Caenorhabditis sp., and the fungus Aspergillus sp. were revealed. For the yeast Saccharomyces sp. only one conservative consensus, /GUA(U,A,C)G(U,A,C), was revealed (E1 = 0.03, E2 = 0.03). Yeast is a very interesting model to use for analysis of molecular mechanisms of splicing. Received: 14 October 1996 / Accepted: 30 January 1997