The problem considered here is that of estimating the law of a discrete variable X=(X1,... ,X I) when the size n of the sample is small compared to the size of the contingency table representing all ...