I have trained different models for prediction bankruptcy 1 year prior bankruptcy, 2 years and 3 years prior and so on. When I use the models on a single sample and I for example get following results:
- $0.4$
- $0.8$
- $0.5$
So can I say that this company goes bankrupt most probably in 2nd year or should i say during 2 years time and with probability 80%?
Or should I calculate valid distribution so $(0.4 + 0.8 + 0.5 = 1.7)$ and
- $0.24$
- $0.47$
- $0.29$
and now I can say that the bankruptcy happens in 2nd year with probability of 47%?
Thank you very much!
Based on the numbers you have indicated ($.4$, $.8$, $.5$), it seems your model is pretty flawed (else, please state precisely what it is your model is calculating). For, it calculates the probability of bankruptcy within the next three years to be $1.7$, which is larger than $100 \%$. Maybe you just made these numbers up, and your model did not actually return them.
As long as you use mathematically precise wording, you can present whatever statistics you want, depending on what you want to express.
Some random notes:
If there is a greater than $.5$ probability that a company will go bankrupt in the $n$th year (for example the second year) you can safely say the company will most likely go bankrupt in this year rather than any other year, based on your model.
If you have only calculated probabilities for one year, two years, and three years, and they come out to $.1, .2, .1$, you cannot necessarily say the company will most likely go bankrupt in the second year. For, it could turn out that the probability for the fourth year is $.4$, higher than the other three.
In the above case, if you wanted to add them up $(.1 + .2 + .1 = .4)$ and then say that the company will go bankrupt in the second year with probability $.5$, you must say it this way: "Given that the company goes bankrupt in the next three years, the company will go bankrupt in the second year with probability $.5$."