Analyzing data in SPSS using ROC Curve For categorical variables (nominal)

Question

I use SPSS v25 to build ROC. I have DataSet with the following data:

Case# Dosage  Result
1       DosagA  healthy
2       DosagA  sick
3       DosagB  sick
4       DosageC healthy
....

To analyse Using ROC, I encoded Result as: Healty =1, sick =0

Case# Dosage  Result
1       DosagA  1
2       DosagA  0
2       DosagB  0
4       DosageC 1
....

When trying to build ROC with: Test Variable= result, State variable= Dosage

I get error message:

String Variable are not allowed in the list

Do I have to Encode Dosage with numeric values like:

Case# Dosage  Result
1       1  1
2       1  0
2       2  0
4       3  1
...

Or What is the best solution for ROC Curve using categorical variables (nominal)?

Answer 1

The prediction needs to be numeric. The ROC curve gives sensitivity vs. 1 - specificity for different cut points on a predictor, whether that's a single predictor or a score based on something like a logistic regression.