getting error Error using ==> ProbDistUnivParam.fit at 98 FREQUENCY values must be non-negative integers.

 Error using ==> ProbDistUnivParam.fit at 98
FREQUENCY values must be non-negative integers.

Try to test hypothesis Goodness of fit while trying Poisson dist and getting error Error using ==> ProbDistUnivParam.fit at 98 FREQUENCY values must be non-negative integers.

Please guide

[dpb]

Well, pretty clear error message. Whatever you used for the frequency counts as the input aren't counts (or integers, anyways).

W/o seeing anything else, what else can be said?

[amberly hadden]

bins = 2:4; obsCounts = [2.1 2.2 2.3 2.4 2.5 2.6 2.7 2.8 2.9]; n = sum(obsCounts); pd = fitdist(bins','Poisson','Frequency',obsCounts'); expCounts = n*pdf(pd,bins); I cannot see

Answer 1

[dpb]

obsCounts = [2.1 2.2 2.3 2.4 2.5 2.6 2.7 2.8 2.9];
...
pd = fitdist(bins','Poisson','Frequency',obsCounts');
...

obsCounts values surely don't look like integers to me...

[Image Analyst]

Perhaps he meant to take the histogram first

histogramCounts = histc(obsCounts, 2:4);

One thing is clear though, he hasn't read this http://www.mathworks.com/matlabcentral/answers/13205-tutorial-how-to-format-your-question-with-markup yet.

[amberly hadden]

I have redone it but have got same problem even I multiplied ObsCount with 10 and proceed but have got error again. Thanks anyways

[dpb]

Don't know where the numbers came from, but surely don't look like Poisson-distributed counts even if make integer. Just to solve the non-integer problem, try

round(obsCounts*10)

but still looks like uniform, not Poisson.

Think we need far more background about the problem to have any rational input.

[amberly hadden]

this is what I have to figure out if this small data is having normal,poison or etc distribution. Well I'm trying different techniques hopefully i will get some results which would agree with any of hypothesis.(that is data has ...... distribution)

[dpb]

More to the point is where did the data come from and what do they represent? They certainly aren't counts from a Poisson; they appear to be simply a linear sequence. W/O any context, the question makes no sense.

[amberly hadden]

this is a very complex data and a very important data set (though just a tiny part of whole data set) if you see no linear line is going to fit on it unless you use predicted values

[dpb]

...this is a very complex data and a very important data set

Which still tells us precisely -- nothing useful.