Isn't that a classification problem? Input data is the BMI, right? Output data should be the state? So once you have a model you can predict the state from a childs BMI?
Or do you want to find a set of states, which have similar BMI distributions?
In this case I think there are additional prior assumptions needed, eg. how the BMI is distributed (I would assume normal distributed, but I could imagine cases with a bimodal distribution as well).
Could you explain in more detail, what the call should be? I didn't understood that yet.
i need to want a set of stats which have similar BMI distributions.
actually i want to find clusters of stats where the children of relative Same pattern of BMI could grouped. so that analysis of each cluster could be done separately .
Well, an inspection of the data has shown that under the assumption that the BMI is identical distributed in different states (this of course needs to be verified), a new artificial data set could be constructed, describing for each state the BMI distribution by its different moments. With this new data set, the states could be clustered according to similar BMI distributions.