Suppose I have this small data set and it is required to fit a compound poisson distribution based on the data [with respect to the the real blood corpuscular (x) per cell] available.

\begin{array}{| c | c |} \hline \\\text{x} & \text{Number of Cells} \\ \\\hline 1& f_1 \\\\\hline 2& f_2\\\\\hline 3& f_3\\\\\hline 4& f_4\\\\\hline 5 & f_{5}\\\\\hline \end{array}

Also $\sum_{i=1}^5 f_i=N$

There is a result that if a random variable $\text{X}\sim \mathcal{Poisson}(\lambda)$ then the compound poisson distribution of $\text{X}\sim \mathcal{Neg \ Bin }(q,p)$ .

Suppose, if it is assumed that $\lambda = \frac{\sum\mathbb{f_ix_i}}{\sum\mathbb{f_i}}$, then I need to calculate $\text{g}(\lambda) = \begin{cases} \frac{ae^{-a\lambda}\lambda^{v-1}}{\Gamma(V)} & \mathrm{ } \lambda>0, a>0, V>0 \\ 0 & \mathrm{ } \lambda \leq 0 \end{cases}.$, where $a$ & $V$ should be determined. Once they are determined then I can find out $$\mathbb{P}\mathrm{(X=x)=\binom{-V}{x}} \textrm{p^V q^x}, \ \ \textrm{ where } \mathrm{p=\frac{a}{1+a}, \ q=1-p,} \ \ \textrm{ x=0,1,2,..,5}$$

Now applying newton-raphson iteration the expected frequencies $e_i$'s can be computed.

Now here are my questions-

  1. how to evaluate $a$ & $V$ while finding out $\text{g}(\lambda)$ on the basis of assumed value of $\lambda$?
  1. Is the column - 'Number of Cells' alone insufficient to fit a compound poisson distribution?
  1. if my approach is incorrect then what is the example or the general structure to fit a compound poisson distribution based on such small data?

Here is one partially related problem but yet it is different. Any help is highly valued and appreciated.


Your Answer

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

Browse other questions tagged or ask your own question.