I am a little lost on how to do a multinomial testing with 3 outcomes, where i want to calculate the probability of success based on following data:

Success: $p^1 = 1/8, y^1=28$

Failure: $p^2=1/8, y^2=10$

Unknown: $p^3=3/4, y^3=22$

$N=60, n=28, \alpha=0.05$

And the calculation:

$f(N,n,p_1^{y_1},...,p_k^{y_k})=\frac{N!}{2^N (n! (N-n)!)}\cdot p_1^{y_1}\cdot ...\cdot p_k^{y_k} = 3.54957*10^{-48}$

Obviously the result allows me to reject the null hypothesis hence its lower than my $\alpha$, but i'm not certain if i am doing this correctly?


1 Answer 1


Since thr probabilities under the null hypothesis is prespecified, this is a simple chisquared test. In R:

yourdata <- c(28, 10, 22)
yourp   <- c(1/8, 1/8, 3/4)
chisq.test(yourdata, p=yourp)

    Chi-squared test for given probabilities

data:  yourdata
X-squared = 68.622, df = 2, p-value = 1.256e-15

The calculation in the post seems to be only the multinomial probability for the exact outcome. That is not enough to define a p-value, you need to define a test function.

  • 1
    $\begingroup$ It is difficult to determine what the OP is trying to ask: the notation is overloaded and obscure and the test is not described. The calculation in the question is neither a Binomial nor a Multinomial probability. (Under your assumptions, the Multinomial probability is $\binom{60}{28,10,22}(1/8)^{28}(1/8)^{10}(3/4)^{22}\approx 5.75\times 10^{-13},$ far greater than the number given in the question.) $\endgroup$
    – whuber
    Dec 19, 2022 at 19:53

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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