Dillon, W., and Goldstein, M. (1984). Multivariate Analysis. New York: Wiley.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....system as a kind of pre processing tool that transforms the regression problem into a classification one before feeding it into a classification system. We have tested RECLA in several regression domains with three different classification systems : C4.5 [12] CN2 [3] and a linear discriminant [4, 6]. The results of our experiments show the validity of our search based approach and the gains in accuracy obtained by adding misclassification costs to classification algorithms. In the next section we outline the steps necessary to use classification algorithms in regression problems. Section 3 ....

....best for a classification problem. Equal width intervals (EW) The range of values is divided into N equal width intervals. K means clustering (KM) In this method the goal is to build N intervals that minimize the sum of the distances of each element of an interval to its gravity center [4]. This method starts with the EW approximation and then moves the elements of each interval to contiguous intervals if these changes reduce the referred sum of distances. This is the more sophisticated method and it seems to be the most coherent with the way we make predictions with the learned ....

Dillon,W. and Goldstein,M., Multivariate Analysis. John Wiley & Sons, Inc, 1984

.... classification rule becomes j = arg min j d j (x) for j = 1; 2; Delta Delta Delta ; k (13) with d j (x) x Gamma m j ) T Sigma Gamma1 j (x Gamma m j ) ln j Sigma j j Gamma 2 ln ff j (14) This equation is often called the discriminant score for the jth class in the literature [25]. Furthermore, if the prior density ff j is the same for all classes (an equal sample number in each class) it becomes discriminant function when omitting the term 2 ln ff j . If a pooled covariance matrix is used, it is called linear discriminant analysis (LDA) which was used by Munson and ....

W. R. Dillon and M. Goldstein, Multivariate Analysis, Wiley, New York, 1984.

....u, and the cluster dendrogram t which corresponds to the chosen number of clusters K. For example, a dendrogram (or in a similar way a cluster dendrogram) can be prepared by f = tree (u 0.0 center xaxis) show(f s2d) 1 2 3 6 7 8 9 5 4 0.0 2.0 4.0 6.0 8.0 SS[ 1] 0.0 0.2 0.4 0.6 0.8 1. 0 SS[ 2] ( 10 1 ) Figure 14: Dendrogram of the columns of Table 1. The Ward method based on the chi square distance is applied. where the input vector u is the result of the command agglom described above. Figure 13 shows the dendrogram of the Ward s clustering of the rows of the contingency table 1. The ....

Dillon, W. R. and Goldstein, M. (1984). Multivariate Analysis, Wiley, New York.

.... of classes) is approximately distributed as a 2 random variable with (n Gamma p) C Gamma p Gamma 1) degrees of freedom under the null hypothesis that the subspace spanned by the right eigenvectors of W Gamma1 B corresponding to the eigenvalues j do not contribute to any discrimination [33]. Starting from the smallest eigenvalue, a hypothesis test is performed, and discriminant directions are cumulatively rejected till the point when the null hypothesis H 0 is rejected. Since speech feature vectors are not independent, if N is chosen as the total number of data frames, this analysis ....

W. R. Dillon and M. Goldstein, Multivariate Analysis. John Wiley and Sons, 1984.

....measures will likely provide more information, the first few principal components will capture a large proportion of the software measurement variance. Thus, principal components are favored as we can avoid the model selection problems that stem from correlations among the independent variables [2]. For more information about principal components analysis, refer to [2] and to Appendix A. The classification (dependent) variable, testability, is based on the probability that a series of tests will detect a fault given that a fault in the module exists. The modules are divided into two groups ....

....components will capture a large proportion of the software measurement variance. Thus, principal components are favored as we can avoid the model selection problems that stem from correlations among the independent variables [2] For more information about principal components analysis, refer to [2], and to Appendix A. The classification (dependent) variable, testability, is based on the probability that a series of tests will detect a fault given that a fault in the module exists. The modules are divided into two groups based on a cutoff value. Modules exceeding the cutoff point are ....

[Article contains additional citation context not shown here]

W. R. Dillon and M. Goldstein. Multivariate Analysis. John Wiley and Sons, New York, 1984.

.... control to participate in multiple clusters, it also adds information from the Euclidean metric to characterize semantic distance. control change power influence time Figure 2: Euclidean model of a portion of the noun based corpus. The first method attempted will be Multi Dimensional Scaling [3]. Given a dimensionality N and a set of interpoint distances, mds produces rotationally invariant Euclidean N space representations. However, mds uses only pair wise information about points. A more sophisticated approach which can represent more context dependence is the MAximum Likelihood ....

Dillon, W.R., and M. Goldsten: (1984) Multivariate Analysis, Wiley, 1984.

....a clustering that is guaranteed to be optimal in terms of the chosen quality measure is an infeasible task, so the available distance based clustering algorithms use heuristic strategies. One of the most basic distance based clustering algorithms is Average Linkage bottom up clustering (c.f. [9]) This method starts out with the list of cases as one element clusters, and iteratively groups the two most similar clusters into a new cluster. Removing both and inserting their union instead, we do this all over again until eventually the cluster list contains only a single cluster (see Figure ....

W. Dillon and M. Goldstein. Multivariate analysis, pages 157--208. John Wiley & Sons, Inc., 1984.

....elements. Equal width intervals (EW) The original range of values is divided into N intervals with the same range. K means clustering (KM) In this method we try to build N intervals that minimize the sum of the distances of each element of an interval to the interval s gravity center 4 [3]. This is basically the P class method that is given in [20] This method starts with the EP approximation but then tries to move the elements of each interval to contiguous intervals whenever these changes reduce the referred sum of distances. To better illustrate these strategies we show how ....

. Dillon,W. and Goldstein,M. (1984) : Multivariate Analysis.. John Wiley & Sons, Inc.

....of Brodatz textures in training will construct a discriminant function general enough to discriminate between new and unknown textures. From the feature sets defined above, we use the Fisher Discriminant Analysis technique to achieve the maximum average separation among different texture classes [22]. The Mahalanobis distance (EQ 1) in the transformed feature space (i.e. after Fisher Discriminant Analysis) was used to measure the similarity between textures. In ordinary classification of textures or comparisons of many textures, the relative ranking of the Mahalanobis distances is used to ....

William R. Dillon and M. Goldstein, Multivariate Analysis, John Wiley & Sons, 1984.

....a clustering that is guaranteed to be optimal in terms of the chosen quality measure is an infeasible task, so the available distance based clustering algorithms use heuristic strategies. One of the most basic distance based clustering algorithms is Average Linkage bottom up clustering (c.f. [9]) This method starts out with the list of cases as one element clusters, and iteratively groups the two most similar clusters into a new cluster. Removing both and inserting their union instead, we do this all over again until eventually the cluster list contains only a single cluster (see Figure ....

W. Dillon and M. Goldstein. Multivariate analysis, pages 157--208. John Wiley & Sons, Inc., 1984.

....of their correlation matrices have been evaluated and sorted giving the principal structure within the eigenvectors of the first z largest eigenvalues. The first 50 largest eigenvalues of 400 of fig. 1 are shown in fig. 3. This analysis of eigenvectors is called principal component analysis, PCA, [7, 4]. The noise and contamination of figures 1 and 2 can successfully be eliminated by transforming the data Z as follows Z S;z = p 1 ; p 2 ; p z ) p 1 ; p 2 ; p z ) T Z S : For all coefficients z S;z;i of the matrix Z S;z a transformation from the centered and standardized ....

W. R. DILLON AND M. GOLDSTEIN, Multivariate Analysis, Wiley, 1984.

No context found.

Dillon, W., and Goldstein, M. (1984). Multivariate Analysis. New York: Wiley.

No context found.

W. R. Dillon and M. Goldstein, Multivariate Analysis, Wiley, New York, 1984.

No context found.

W. Dillon and M. Goldstein. Multivariate analysis, pages 157#208. John Wiley & Sons, Inc., 1984.

No context found.

Dillon,W. and Goldstein,M. (1984) : Multivariate Analysis. John Wiley & Sons, Inc.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC