Source author record

Necla Gunduz

Explore this author’s publications, collaborators and research topics. This catalog record has not been claimed by its author.

ResearcherUnclaimed source record

Applications Methodology math.ST Statistics Theory

Catalog footprint

Research overview

3works

4topics

1close collaborators

Actions

Is this your work?

Open graph Browse papers

preprint2015arXiv

An Information-Theoretic Alternative to the Cronbach's Alpha Coefficient of Item Reliability

We propose an information-theoretic alternative to the popular Cronbach alpha coefficient of reliability. Particularly suitable for contexts in which instruments are scored on a strictly nonnumeric scale, our proposed index is based on functions of the entropy of the distributions of defined on the sample space of responses. Our reliability index tracks the Cronbach alpha coefficient uniformly while offering several other advantages discussed in great details in this paper.

preprint2015arXiv

Pattern Discovery in Students' Evaluations of Professors: A Statistical Data Mining Approach

The evaluation of instructors by their students has been practiced at most universities for many decades, and there has always been a great interest in a variety of aspects of the evaluations. Are students matured and knowledgeable enough to provide useful and dependable feedback for the improvement of their instructors' teaching skills/abilities? Does the level of difficulty of the course have a strong relationship with the rating the student give an instructor? In this paper, we attempt to answer questions such as these using some state of the art statistical data mining techniques such support vector machines, classification and regression trees, boosting, random forest, factor analysis, kMeans clustering. hierarchical clustering. We explore various aspects of the data from both the supervised and unsupervised learning perspective. The data set analyzed in this paper was collected from a university in Turkey. The application of our techniques to this data reveals some very interesting patterns in the evaluations, like the strong association between the student's seriousness and dedication (measured by attendance) and the kind of scores they tend to assign to their instructors.

preprint2015arXiv

Robust Classification of High Dimension Low Sample Size Data

The robustification of pattern recognition techniques has been the subject of intense research in recent years. Despite the multiplicity of papers on the subject, very few articles have deeply explored the topic of robust classification in the high dimension low sample size context. In this work, we explore and compare the predictive performances of robust classification techniques with a special concentration on robust discriminant analysis and robust PCA applied to a wide variety of large $p$ small $n$ data sets. We also explore the performance of random forest by way of comparing and contrasting the differences single model methods and ensemble methods in this context. Our work reveals that Random Forest, although not inherently designed to be robust to outliers, substantially outperforms the existing techniques specifically designed to achieve robustness. Indeed, random forest emerges as the best predictively on both real life and simulated data.

Inspect adjacent papers, topics, institutions and collaborators without losing the researcher page.

Building this map preview

BZPEER is loading the nearby papers, people, topics and institutions for this page.

Necla Gunduz

Research overview

Is this your work?

3 publications

An Information-Theoretic Alternative to the Cronbach's Alpha Coefficient of Item Reliability

Pattern Discovery in Students' Evaluations of Professors: A Statistical Data Mining Approach

Robust Classification of High Dimension Low Sample Size Data

See the researcher in context

Building this map preview