Can I perform sensitivity analysis, if I don't know expected prediction results?

I.e. I have a model with input parameters and weights.

But I don't know when a prediction should be true and when false.

Can I still perform sensitivity analysis, what kind?

  • $\begingroup$ Hello, I am not sure to understand your question. By sensitivity analysis you mean check the sensitivity of the model parameter with respect to the data on which it was fitted (in which case robust statistics could help)? or the sensitivity of a prediction with respect to explanatory variables ? or something else ? $\endgroup$ – Pohoua Feb 3 at 1:11
  • $\begingroup$ @Pohoua Thing is, I don't know. I was guided to sensitivity analysis, but my project does not have "output labels". That is I don't know what kind of prediction should ring "true prediction" and what should ring "false prediction". Then the question is, what does sensitivity analysis mean, when it's done only on the model without being able to compare to "true predictions". I've speculated that it's about measuring coefficient ranges that produce model(input)=score_i. So that one can understand whether there are more contradictory confs of $(b_1,...b_n)$ coefs that produce score_i. $\endgroup$ – mavavilj Feb 3 at 8:16

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.