HI!
I am working with a model with as output a list of probabilities to attain a result. For example,
- result 1: 0.99
- result 2: 0.76
- ...
- result 10: 0.56
This means that the user has 99% chance to attain result 1, 76% chance to attain result 2, etc. The probabilities of the different results do not sum up to 1 as they have to be viewed separately as they correspond to attain a result within a specified time period which could be different for each result. Now I want to change one of the input parameters slightly by adding a small error and rerun the model 50 times. Every time I would have again an output consisting of a list of 10 probabilities, which will be slightly different from the 'real' probabilities above. For example, I could have an outcome like this
- result 1: 0.97
- result 2: 0.72
- ...
- result 10: 0.47
If I repeat this 50 times I could calculate the average difference between the real probability to attain that result and the incorrect ones after adding an error. But, how can I now test/prove if the average deviation between the real probability and the incorrect ones for the list as a whole or for each result individually is large or small?
I am working with a model with as output a list of probabilities to attain a result. For example,
- result 1: 0.99
- result 2: 0.76
- ...
- result 10: 0.56
This means that the user has 99% chance to attain result 1, 76% chance to attain result 2, etc. The probabilities of the different results do not sum up to 1 as they have to be viewed separately as they correspond to attain a result within a specified time period which could be different for each result. Now I want to change one of the input parameters slightly by adding a small error and rerun the model 50 times. Every time I would have again an output consisting of a list of 10 probabilities, which will be slightly different from the 'real' probabilities above. For example, I could have an outcome like this
- result 1: 0.97
- result 2: 0.72
- ...
- result 10: 0.47
If I repeat this 50 times I could calculate the average difference between the real probability to attain that result and the incorrect ones after adding an error. But, how can I now test/prove if the average deviation between the real probability and the incorrect ones for the list as a whole or for each result individually is large or small?