Notes

  1. Patientid has duplicated numbers. I assume each row represent one independent observation.
  2. What is PBD? I used Vasospasm = 0 as No Vasospasm; Vasospasm <= 1 as Mild Vasospasm; Vasospasm > 1 as severe vasospasm.

Boxplots

ROC curve - to predict Vasospasm (Yes/No)

1. single marker: sFlt1

2. single marker: sEng

3. combined sFlt1 and sEng

4. best cut off

As we can see, the combined marker has lower predictibility than single marker sFlt1.

Best cutoff to predict Vasospasm (Yes/No) is Serum sFlt1 day 6 > 186.45.

## Area under the curve: 0.7628
## sensitivity specificity         ppv         npv 
##   0.7692308   0.7500000   0.7692308   0.7500000

ROC curve - to predict Vasospasm (Mild/Severe)

1. single marker: sFlt1

2. single marker: sEng

3. combined sFlt1 and sEng

4. best cut off

As we can see, the combined marker did not yield higher predictibility than single markers.

Best cutoff to predict Vasospasm (Mild/Severe) is CSF sFlt1 day6 > 136.8.

## Area under the curve: 0.8095
## sensitivity specificity         ppv         npv 
##   1.0000000   0.5833333   0.5833333   1.0000000