Software_index

Software programs available from the Kurt Rossmann Laboratories

Software description

Programs that fit ROC curves and test differences between them

Auxiliary ROC software

Software description

Programs that fit ROC curves and test differences between them

ROCKIT	Brief Description ROCKIT uses maximum likelihood estimation to fit a binormal ROC curve to: Continuously-distributed data (e.g., probability assessment on a 0-100 scale or numerical output from a laboratory measurement like a blood count) AND/OR Ordinal category data (e.g., discrete confidence ratings, where the categories could be: "definitely or almost definitely negative", " probably negative", "possibly positive", "probably positive", and "definitely or almost definitely positive" . ROCKIT also calculates the statistical significance of differences between ROC index estimates and parameters. On the basis of a "bivariate binormal" model, it allows for comparison of 2 paired, partially paired, or unpaired datasets (which would represent, for example, different imaging modalities or diagnostic tests) with regard to: Differences in the binormal ROC curve parameters a and b (related to the difference in the mean and of the standard deviations of the two latest normal distributions used in the fit) Difference in the areas (Az) under the two estimated binormal ROC curves Difference between the two true-positive fractions (TPF)s on the two curves at a selected false-positive fraction (FTP) Platforms: Mac and Windows References: ROCKIT, its usage, and the algorithms it employes Studies that have used ROCKIT ROCKIT REPLACES ROCFIT, LABROC, CORROC2, CLABROC and INDROC. If the functionality of any of these programs is required, ROCKIT should be used instead. None of the older programs will be updated or modified in the future. ROCKIT employs our LABROC5 algorithm, a quasi-maximum likelihood approach, to analyze continuously distributed data. The statistical tests performed by ROCKIT allow the conclusions of a study to be generalized to a population of cases, but not to a population of readers. If generalization to populations of both of readers and cases are required, the program LABMRMC should be used instead.
LABMRMC	Brief Description LABMRMC uses the Dorfman-Berbaum-Metz algorithm to compare multiple treatments (e.g., imagining modalities) by using data from multiple readers and multiple cases. This program employs jackknifing and ANOVA techniques. Platforms: Mac and Windows References: LABMRMC, its usage, and the algorithms that it employs Studies that have used LABMRMC This program allows the conclusions drawn from a study to be generalized to bith a population of readers and a population of cases
LABROC4	Brief Description LABROC4 uses maximum likelihood estimation to fit a single conventional binormal ROC curve to continuously-distributed data (e.g., probability assessment on a 0-100 scale or numerical output from a laboratory measurement such as blood count). This program does not evaluate the statistical significance of the differences between ROC estimates Platforms: program in development, available only through direct request References: LABROC4, its usage, and the algorithms it employs Studies that have used LABROC4 This program uses a true maximum-likelihood approach rather than the quasi-maximum likelihood approach (LABROC5) that is used in ROCKIT. It will be phased out and incorporated into ROCKIT at some point in the near future.
PROPROC	Brief Description PROPROC fits a single ROC curve by maximum-likelihood estimation (given ordinal category data) or by quasi maximum-likelihood estimation (given continuously-distributed data). However, unlike ROCKIT or LABROC4, PROPROC employs a "proper" binormal model that always provides convex fitted curves. Thus, this program allows the fitting of degenerate data sets (datasets which can be fit with vertical and horizontal straight line segments, which are impossible to fit using the conventional binormal model) and data sets that would yield non-monotonic (i.e., non-convex) fits with the conventional binormal model. This program does not provide estimates of standard error and does not evaluate the statistical significance of differences between ROC estimates Platforms: program in development, available only through direct request References PROPROC, its usage, and the algorithms it employs Studies that have used PROPROC This program is unique among our ROC software in the sense that it is the only program that employs the "proper" binormal model

Auxiliary ROC software

PlotROC.xls

Brief Description

This is a Microsoft Excel 5.0 macro sheet which takes the a and b parameter values of the conventional binormal model and plots an ROC curve suitable for presentation and publication.

Platforms: Mac and Windows

It requires Microsoft Excel 5.0

ROCPWR

Brief Description

This program predicts the power of the statistical test that is performed by ROCKIT with fully-paired data. It requires, as input, values of the conventional binormal parameters a and b for each of the two ROC curves (which can be obtained by guesswork or by running ROCKIT or LABROC4 with pilot-study data) and the decision variable correlation parameters (which can be obtained by guesswork or by running ROCKIT or LABROC4 with pilot-study data). ROCPWR outputs a table of estimated statistical power for a variety of case sample sizes.

Platforms: Windows

References

ROCPWR, its usage, and the algorithms it employs

Download software

If you are having trouble downloading the files please check our

Download Troubleshooting Page

Contact roc at bsd uchicago edu with questions/suggestions about this Section

Home | Up | History | Members | Projects | Research Summary | ROC Software | Publications | Topicsandnews | University Links | Contact Us

This site was last updated 05/05/04