I have two multivariate normal distributions (say X and Y) fitted on existing data. With new data I want to compute the probability that sample is from distribution X instead of Y. It can only be one of these two distributions.

Any ideas on how I can go about doing this?

  • 1
    $\begingroup$ To compute such a probability you would need a prior probability on the two. From there it's a simple application of Bayes theorem $\endgroup$ – Glen_b -Reinstate Monica Sep 10 '16 at 1:21

If the density functions of $X$ and $Y$ are $f_X$ and $f_Y$, and the prior probabilities of the data vector $d$ being drawn from distribution $X$ and $Y$ are $p_X$ and $p_Y$ respectively then the probability the data came from distribution $X$ can be found using Bayes theorem:

$$ \frac{f_X(d) p_X}{f_X(d) p_X + f_Y(d) p_Y} . $$


Your Answer

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

Not the answer you're looking for? Browse other questions tagged or ask your own question.