Home   Consulting   About Us   Contact Us   Donate
MedRoc Home   MedRoc ScreenShots   MedRoc Download

MedRoc Features

MedRoc was designed with a set of features that provide maximum flexibility in the kinds of analyses that can be performed, along with an output format that provides convenient choices among the aspects of the output that you wish to examine, save, or print. In addition, the MedRoc Editor provides a means to enter data from within the MedRoc program or to import data from an external file into the program. (The MedRoc Screenshots link will show you examples relevant to some of these features.)

The topics with links shown below list the major features of each of the various aspects of MedRoc. Click a link to move to a different topic.

Go to  General Features   MedRoc Analysis   MedRoc Output   MedRoc Editor

General Features

• The data-editor is integrated into the main window of MedRoc.

• Up to 8 Surrogate (response) variables (columns of data) are allowed and will be analysed for the same set of Gold Standard values.

• ROCs for multiple Surrogates may be shown as separate graphs or all on the same graph.

• Bootstrap analysis of the area under the ROC is used to statistically compare all pairs of multiple surrogate variables.

• Bootstrap analysis will statistically compare the ROC areas for all pairs of designated groups (rows of data) if there is only one surrogate.

• Bayesian probabilities are calculated and displayed for both the Positive Predictive Value (PPV) and Negative Predictive Value (NPV).

• An extensive Help file is included in MedRoc.

Go to  General Features   MedRoc Analysis   MedRoc Output   MedRoc Editor

MedRoc Analysis

• Either the non-parametric or binormal ROC model can be used.

• For the binormal ROC model you can specify the number of maximum-likelihood iterations that are to be run, and the tolerance level for stopping the iterations.

• MedRoc will handle input data that are either discrete or continuous in nature, e.g. rating scale data or blood-pressure readings.

• You can perform ROC analysis by groups of data cases. A separate analysis is run for each group that you designate by using a grouping variable.

• You can specify that ROC curves for multiple groups be shown on the same ROC graph for comparison, or you can ask for a separate graph for each group.

• You can select a particular subset of cases for analysis using logical ANDs and ORs on your variables.

• For data that are already aggregated you can specify a Frequency Variable that contains the frequency of occurrence for each case (but the data do not have to be already aggregated).

• You can specify which of two Gold Standard values is to be considered the "high" value.

• Input data are checked to make sure that they contain no more than two values of the variable you specify as the Gold Standard.

• You can specify whether the high values of the Surrogate Variable (responses) are intended to indicate high or low values of the Gold Standard.

• You can specify prior probabilities for the occurrence of the high value of the Gold Standard. This enables some Bayesian output (see the output section below)

• The costs of false negative and false positive decisions may be specified, as well as benefits of correct decisions. These values are used to calculate expected values for using any particular decision criterion (see output section below). The cost and benefit values are scale-free, i.e. only their relative sizes are used in the calculation of expected values.

Go to  General Features   MedRoc Analysis   MedRoc Output   MedRoc Editor

MedRoc Output

• The output of an ROC analysis is shown in 6 overlapping windows.(See MedRoc Screenshots for an example.) Each window contains data specific to a particular aspect of output. For example, one window shows frequency data and probabilities. Another shows the statistics and confidence intervals associated with the ROC. Another shows the ROC graph(s) for the analysis, etc.

• Any of the 6 output windows can be temporarily closed to reduce visual clutter, and any closed window can be reopened if you wish to view it again.

• The data in a particular output window can be scrolled to show similar data for each the subgroups that you may have specified.

• The data in the 6 output windows can be saved or printed. You can choose which of the windows is to be saved or printed; any or all of them.

• You may choose a default file into which saved output will be directed.

• Saved output from previous runs of MedRoc can be viewed (but not edited) in the MedRoc main output window.

• ROC graphs for separate groups of data will display either individually or all on the same graph depending on your choice. ROC graphs may be saved as bitmap (.bmp) files.

• The output window for ROC statistics shows the value of the area under the ROC curve and various other estimated parameters depending on the kind of analysis model that was used. The standard errors and confidence intervals are shown for all estimated parameters.

• For multiple groups of data, the confidence intervals for the areas under the ROC curves are displayed on a graph so that the groups may be compared to one another. This graph may be saved as a bitmap file.

• For each instance of the Surrogate Variable (responses) a Bayesian probability is shown that is the probability of having a high value of the Gold Standard given that the particular Surrogate Value is used as a decision criterion. This probability depends on the prior probability that you specify as the unconditional probability of a high value of the Gold Standard.

• For each instance of the Surrogate Variable (responses) an expected value is shown for using that particular Surrogate Value as a decision criterion. The expected values depend on the prior probability of a high value of the Gold Standard and on the costs and benefits you specified.