A comparison of discriminant procedures for binary variables

Asparoukhov, Ognian K.; Krzanowski, W. J.

doi:10.1016/s0167-9473(01)00032-9

Cited by 31 publications

(12 citation statements)

References 30 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This result is also normally observed in many other research fields. For example, in the comparison of 13 discriminant procedures for binary data, Asparoukhov and Krzanowski (2001) showed that the traditional statistical classifiers did not cope well with sparse binary data, while neural networks were much better in such circumstances. In the prediction of stone disease in genetic polymorphisms, the neural networks provided better predictive power than did classical discriminant analysis (Chiang et al, 2003).…”

Section: Discussionmentioning

confidence: 99%

Hazard rating of pine trees from a forest insect pest using artificial neural networks

Park

Chung

2006

Forest Ecology and Management

View full text Add to dashboard Cite

Section: Discussionmentioning

confidence: 99%

Hazard rating of pine trees from a forest insect pest using artificial neural networks

Park

Chung

2006

Forest Ecology and Management

View full text Add to dashboard Cite

“…Furthermore, as shown in many studies that have been conducted, the losses incurred by LDF under nonoptimal conditions compared with other procedures are small enough not to be of any practical importance. 17 Applying the results of Su and Liu 13 to our binary variables, the optimal linear combination is given by the coefficients:…”

Section: Heritability and Pafmentioning

confidence: 99%

Evaluation of the discriminative accuracy of genomic profiling in the prediction of common complex diseases

2009

View full text Add to dashboard Cite

Genetic testing for susceptibility to common diseases based on a combination of genetic markers may be needed because the effect size associated with each genetic marker is small. Whether or not a genome profile based on a combination of markers could yield a useful test can be evaluated by assessing the discriminative accuracy. The authors present a simple method to calculate the clinical discriminative accuracy of a genomic profile when the relative risk and genotype frequency of each genotype are known. In addition, the clinical discriminative accuracy of a genetic test is presented for given values of the heritability and prevalence of the disease and for the population-attributable fraction of the combined genetic markers. For given values of relative risk and genotype frequency, the discriminative accuracy increases with increasing heritability but declines with increasing prevalence of the disease. For a given value of population-attributable fraction, the discriminative accuracy increases with increasing relative risks, but declines with increasing genotype frequency. On the basis of populationattributable fraction and estimates of heritability of disease, the number of risk genotypes required to have a reasonable clinical discriminative accuracy is much higher than the genome profiles available at present.

show abstract

“…In order to gauge the overall performance of the stand-alone and hybrid models, we computed empirical integrated rank (EIR), as suggested by Asparoukhov and Krzanowski (2001), as follows:…”

Section: Performance Evaluations Using Eirmentioning

confidence: 99%

Leveraging the Strengths of Choice Models and Neural Networks: A Multiproduct Comparative Analysis*

Papatla¹,

Zahedi²,

Zekić–Sušac³

2002

Decision Sciences

View full text Add to dashboard Cite

Choice models and neural networks are two approaches used in modeling selection decisions. Defining model performance as the out-of-sample prediction power of a model, we test two hypotheses: (i) choice models and neural network models are equal in performance, and (ii) hybrid models consisting of a combination of choice and neural network models perform better than each stand-alone model. We perform statistical tests for two classes of linear and nonlinear hybrid models and compute the empirical integrated rank (EIR) indices to compare the overall performances of the models.We test the above hypotheses by using data for various brand and store choices for three consumer products. Extensive jackknifing and out-of-sample tests for four different model specifications are applied for increasing the external validity of the results. Our results show that using neural networks has a higher probability of resulting in a better performance. Our findings also indicate that hybrid models outperform stand-alone models, in that using hybrid models guarantee overall results equal or better than the two stand-alone models. The improvement is particularly significant in cases where neither of the two stand-alone models is very accurate in prediction, indicating that the proposed hybrid models may capture aspects of predictive accuracy that neither stand-alone model is capable of on their own. Our results are particularly important in brand management and customer relationship management, indicating that multiple technologies and mixture of technologies may yield more accurate and reliable outcomes than individual ones.

show abstract

A comparison of discriminant procedures for binary variables

Cited by 31 publications

References 30 publications

Hazard rating of pine trees from a forest insect pest using artificial neural networks

Hazard rating of pine trees from a forest insect pest using artificial neural networks

Evaluation of the discriminative accuracy of genomic profiling in the prediction of common complex diseases

Leveraging the Strengths of Choice Models and Neural Networks: A Multiproduct Comparative Analysis*

Contact Info

Product

Resources

About