CODEWITHSUNDEEP
spssroc
Ana
Ana

Reputation: 97

AUC values from ROC curves show discrepancies (spss)

I got two different values for AUC when calculating ROC curves in SPSS (version 24). I have a dataset of 75 samples and 10 variables (ΔCt) for each sample.

First I did ROC analysis for 6 variables.

Code:

ROC ΔCt_1 ΔCt_2 ΔCt_3 ΔCt_4 
    ΔCt_5 ΔCt_6 BY Label (1) 
  /PLOT=CURVE(REFERENCE) 
  /PRINT=SE COORDINATES 
  /CRITERIA=CUTOFF(INCLUDE) TESTPOS(SMALL) DISTRIBUTION(FREE) CI(95) 
  /MISSING=EXCLUDE

Results:

Results_6variables

Then I repeated the analysis for all 10 variables.

Code:

ROC ΔCt_1 ΔCt_2 ΔCt_3 ΔCt_4 
    ΔCt_5 ΔCt_6 ΔCt_7 ΔCt_8 
    ΔCt_9 ΔCt_10 BY Label (1) 
  /PLOT=CURVE(REFERENCE) 
  /PRINT=SE COORDINATES 
  /CRITERIA=CUTOFF(INCLUDE) TESTPOS(SMALL) DISTRIBUTION(FREE) CI(95) 
  /MISSING=EXCLUDE

Results:

Results_10variables

Visually, the curves are the same, but there could be minor differences that are not visible to my eyes. I would like to understand why the same data (for the 6 variables analysed in both calculations) provides different results.

Thank you very much for your help.

Best, Ana

Upvotes: 1

Views: 388

Answers (1)

Rick Marcantonio
Rick Marcantonio

Reputation: 96

Ana - ROC uses LISTWISE deletion to make sure that each curve is comparable to (i.e., uses the exact same observations as) the others.

Upvotes: 1

Related Questions