1

how to find probability cut off in logistic regression. by considering the Sensitivity, specificity and accuracy on different cut off values.

cut off                      0.3    0.35    0.4      0.45      0.5   0.55
Sensitivity                 66.93%  57.48%  44.09%  30.71%  23.62%  18.90%
Specificity                 58.97%  72.16%  82.05%  87.18%  93.04%  95.24%
Accuracy                    61.50%  67.50%  70.00%  69.25%  71.00%  71.00%

In this case which cut off to choose.

Sven Hohenstein
  • 6,285
  • 25
  • 30
  • 39
rijin
  • 11
  • 1
  • 0.3 0.35 0.4 0.45 0.5 0.55 Sensitivity 66.93% 66.93% 57.48% 44.09% 30.71% 23.62% 18.90% Specificity 58.97% 58.97% 72.16% 82.05% 87.18% 93.04% 95.24% Accuracy 61.50% 61.50% 67.50% 70.00% 69.25% 71.00% 71.00% – rijin Jul 06 '17 at 08:27
  • 2
    The choice of cut-off is based on what you what to achieve using this logistic regression model. Although this is oversimplifying things in a major fashion, I was taught this handy mnemonic: **Sp**in rules in, **Sn**out rules out, where sp stands for specificity and sn for sensitivity. Most importantly however, if you could add context about the goal of your study/research, and some information on what is in the model, a more applicable answer might be possible. – IWS Jul 06 '17 at 08:33
  • One seeks a cutoff when one does not understand optimal decision making. – Frank Harrell Jul 06 '17 at 11:16

0 Answers0