This question is from Robert Hogg's Introduction to Mathematical Statistic 6th version exercise 8.5.2 page 461. The question is:

Let $X_1, X_2,...,X_{10}$ be a random sample of size 10 from a Poisson distribution with parameter $\theta$. Let $L(\theta)$ be the joint pdf of $X_1, X_2,...,X_{10}$. The problem is to test $H_0:\theta=\frac{1}{2}$ against $H_1:\theta=1$

(a) Show that $L(\frac{1}{2})/L(1) \le k$ is equivalent to $y=\sum_1^n x_i\ge c$

(b)In order to make $\alpha=0.05$, show that $H_0$ is rejected if y>9 and, if y=9, reject $H_0$ with probability $\frac{1}{2}$(using some auxiliary random experiment)

The solution says (a) and (b) are straightforward, I think (a) is simple, but I cannot see straightforward at (b)

My some thoughts on (b): since $X_1,...,X_{10}$ are random sample for a Poisson so $y=\sum_1^{10} x_i\sim Poisson (5) $ with $H_0$ and $y=\sum_1^{10} x_i\sim Poisson (10) $ under $H_1$

Under $H_0:$

$ y>9, \Rightarrow p=1-ppois(9,5)=0.03182806 $

$ y=9, \Rightarrow p=dpois(9,5)= 0.03626558 $

They are all less than $\alpha=0.05$ what is the meaning of "reject $H_0$ with probability $\frac{1}{2}$"?

Thank you very much.

  • 2
    $\begingroup$ You need to finish the calculation! What is ` (1-ppois(9,5)) + dpois(9,5)/2` equal to? (Incidentally, the probability "1/2" in (b) appears to be an approximation to a number close to $0.5010796$.) $\endgroup$
    – whuber
    Jun 24 '16 at 13:49
  • $\begingroup$ Thank you very much whuber. Now I get the solution. I will write down the straightforward solution which is not straightforward for me at all. $\endgroup$
    – Deep North
    Jun 25 '16 at 14:03

Since the Poisson is a discrete distribution we may not get an significant level at 0.05 exactly we may need an Auxiliary random experiment to make it exact 0.05. we do this way:

$(1-ppois(9,5)) + dpois(9,5)*p=0.05\\\Rightarrow p=0.5010796$


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.