Accumetric Test Performance Analysis software, receiver operating characteristic analysis, nonparametric, sensitivity, specificity, likelihood ratio, predictive value, kappa, weighted kappa, intraclass correlation

What sort of data does Accumetric need?

Accumetric uses comma-separated values ("csv") files, which can be created within Accumetric itself or with any text editor, word processor or spreadsheet. Most spreadsheets, for example, can export data to comma-separated values files.

Data files for analysises of reliability must include the name of the test being analyzed, the name of the condition it is intended to detect, a line for consistency with files for analysis of validity (for reliability files this line can be blank, but must be present), below, and, for each subject, a case number, an optional unique identifier (such as study subject number or subject initials, and an integer or real number test score or value for each subject.

Here is an example of a data file for an analysis of reliability:

Depression scale
Depression
na
1, Ms. HB, 5
2, Mr. YJ, 3
3, Ms. GN, 2
4, Mr. KS, 3
5, Ms. DL, 4

Data files for analyses of validity, including ROC analysis, must include the name of the test being analyzed, the name of the condition it is intended to detect, an indication, either "direct" or "inverse", of whether test scores vary directly or inversely with the condition being detected, and, for each subject, a case number, an optional unique identifier (such as study subject number or subject initials, an integer or real number test score or value for each subject and an independently-derived indication ("gold standard") of whether or not the subject actually has the "disease" that the test is meant to detect, expressed as a "yes" or "no", "y" or "n", "1" or "0", whichever you prefer.

In addition, for ROC analysis, there must be at least 5 score levels so that a reasonable number of data points will appear on the curve, but this is not necessary for other validity analyses.

Here is an example of a data file for ROC analysis:

Depression scale
Depression
direct
1, Ms. HB, 5, yes
2, Mr. YJ, 3, no
3, Ms. GN, 2, no
4, Mr. KS, 3, yes
5, Ms. DL, 4, yes

Finally, you do not need to use Unique Identifiers if you do not wish, as long as you hold their place with a comma. If you create a data file within Accumetric, this place will be held automatically. If you use a spreadsheet, for example, you would leave the second column of data blank. Your data would look like this:

Depression scale
Depression
inverse
1, , 5, yes
2, , 3, no
3, , 2, no
4, , 3, yes
5, , 4, yes