A triptych on three continuous analogs of the Poisson, binomial, and negative binomial distributions (Q6175187): Difference between revisions

Many continuous statistical distributions admit their discrete counterparts. For example, for a normalized positive convergent series belonging to the hypergeometric type, the successive terms there can be looked upon as forming a discrete distribution. Various properties of this discrete distribution can be studied by using the properties of the corresponding hypergeometric series of a positive real scalar variable. If the general term in a positive normalized convergent power series for a continuous variable \(x>0\) is \(x_{(r)}=a_rx^r/r!\), where \(a_r\) is the coefficient, then for a fixed \(r\) the probability that \(x\ge r\) is the same as the probability for the corresponding discrete variable \(n\ge r\) for every \(r\), taking the probability that \(n=r\) is \(x_{(r)}\) where the discrete variable \(n\) is defined as \(n=[x]\), where \([x]\) is the largest integer less than or equal to \(x\). For deriving this survival probability, we sum up the terms in the discrete case and we integrate out the underlying function in the continuous case. Thus, going from a power series for a continuous variable to the corresponding discrete counterpart is a simple process. Is it also possible to obtain the continuous analogue or analogues from a given discrete probability law? This aspect is examined in this paper for the discrete models such as the Poisson law, binomial law and the negative binomial law. The continuous analogues may be called continuous Poisson, continuous binomial etc. Methodologies are provided for constructing such continuous analogues. Potentials of applying such continuous analogues to various problems is also discussed in this paper.