ROC Analysis Educational Worksheet
This is a simple Excel spreadsheet to demonstrate the metrics of detection performance (sensitivity, specificity, etc) and show how these metrics contribute to receiver operating characteristic curves.
Spreasheet contains two worksheets: "ASSIGNMENT"(for educational problems below) and "DynamicTestData" (for generating custom data and testing more scenarios):
Instructions: Spreasheet contains the test results for a hypothetical imaging study of a tracer. When injected in the body, the tracer collects in tumors and background tissue. Higher uptake is associated with malignant lesions, but the optimal uptake level to detect malignancy is under investigation in this study. This study includes imaging tests on over 150 cases where knowledge of malignancy is known with roughly 1/2 of the cases with malignancy and 1/2 without malignancy.
On worksheet, "ASSIGNMENT", Complete the following problems:
1) Calculate Sensitivity, Specificity, Accuracy for Single Test Threshold of 38 (Calculations can be performed inside excel; Complete the yellow cells with appropriate formulas)
2) If you wanted to be sure to detect all disease, what is the highest threshold you could use according to this data?
3) What is the threshold that leads to almost identical sensitivity and specificity? Why?
4) Calculate the Area Under the ROC and SNR below. (Hint: Can perform numerical integration; averaging of TPF to get AROC)
Author: Adam Alessio, https://www.egr.msu.edu/~aalessio/
Last Revised: 30-Sept-2019