# Thread: How can I find the distribution of a series of numbers?

1. ## How can I find the distribution of a series of numbers?

Suppose I have some numbers say x1,x2,x3,....xn . Then is there any way that I can find to which distribution these numbers belong to? if there is some method please help me. I need to know this desperately. Thanks.

2. ## Re: How can I find the distribution of a series of numbers?

Originally Posted by abhishekxi
Suppose I have some numbers say x1,x2,x3,....xn . Then is there any way that I can find to which distribution these numbers belong to? if there is some method please help me.
I not sure what this question means.
But any finite set of numbers can be sorted and then counted.

3. ## Re: How can I find the distribution of a series of numbers?

suppose say..numbers 10 4 8 6 14 24 38..some numbers..what the question is..can we find to which distribution these numbers belong to..i mean can we find..whether they belong to uniform distribution, poisson distribution, normal distribution..or some other..is there anyway..??

BTW, what do you mean by sort and count? could you explain me that please..

4. ## Re: How can I find the distribution of a series of numbers?

You are looking for a distribution fitting software. Once you have a vector of N values (the larger, the better), you may send these data to the program, and it will automatically find the distribution that fits best the observed data (normal, gamma, chi-squared, weibull, among others). The parameters associated with each distribution are also provided along with the goodness-of-fit of each tested distribution.

Check "easyfit" on google. This is a good package for automated fitting. There are several others, nonetheless.

5. ## Re: How can I find the distribution of a series of numbers?

Alonso thank you very much. But I need to know the whole procedure of solving the problem. Actually, I need to write a program to the process. Thanks and if you know something about the process please help me out.

6. ## Re: How can I find the distribution of a series of numbers?

Things may be a lot complicated depending on the distribution you are working with, but the following link gives what you are looking for:

Let me google that for you

7. ## Re: How can I find the distribution of a series of numbers?

Actually I would not know about the distribution. I need to fit some distribution that best fits my data. I need some procedure for it. All my efforts are going in vain.

8. ## Re: How can I find the distribution of a series of numbers?

Yes, I understand your problem. check that link. Once you have the data at hand, you may follow some tutorials that are found there (there's a good one with examples in R).

You pick up a distribution, say, normal, then test its fit. If it does to fit well (using appropriate methods - explained there), you move to the next one, say, gamma. It is a trial and error approach, if you want to perform it manually. that's why you were told about the software - as pointed out above.

9. ## Re: How can I find the distribution of a series of numbers?

Originally Posted by abhishekxi
Suppose I have some numbers say x1,x2,x3,....xn . Then is there any way that I can find to which distribution these numbers belong to? if there is some method please help me. I need to know this desperately. Thanks.
What you ask is impossible, what you need to know is a plausible family of distributions that you apriori wish to consider. Then you test the observed data against the family possibly choosing the one which gives the maximum likelihooh.

So the first think you need to ask is what process is giving rise to this data, and in light of this what are reasonable type of distribution to consider for representing this data.

CB

10. ## Re: How can I find the distribution of a series of numbers?

the data is demand for a product on different days..i think poisson distribution fits well for it..but I don't know..the demand on one particular day is independent of other..that made me inclined towards poisson distribution..
if anyone could help me. One thing, can I assume different distributions for same data and choose the best fit by applying goodness of fit tests? Can I do that?

-Thanks and Regards.