Conditions for approximation of the bias and mean squared error of a sample mean under nonresponse (Q1202307)

Consider a sequence of populations \(\{U_ t,\;t=1,2,\dots\}\) where \(U_ t\) has \(N_ t\) distinct units. The \(i\)-th unit in \(U_ t\) has characteristic \(y_{ti}\) which is treated as fixed while randomness is induced only through sampling and a nonresponse mechanism. We wish to estimate \(\overline Y_ t=N^{-1}_ t\sum_{i\in U_ t} y_{ti}\). Consider the following conditions: \((\text{C}_ 1)\) For each \(t\), a simple random sample \(s_ t\) of size \(n_ t\) is drawn without replacement from \(U_ t\), and \(n_ t\) and \(N_ t\) are increasing in \(t\). \((\text{C}_ 2)\) for each \(t\), \(\{\beta_{ti}, i\in U_ t\}\) is a set of independent Bernoulli random variables with response probability for unit \(i\) being \(\eta_{ti}\). \((\text{C}_ 3)\) The first and second population moments of \(y_{ti}\) are bounded uniformly in \(t\). \((\text{C}_ 4)\) The first four population moments of \(y_{ti}\) are bounded uniformly in \(t\). Using \(\text{C}_ 1\), \(\text{C}_ 2\), \(\text{C}_ 3\) or \(\text{C}_ 1\), \(\text{C}_ 2\), \(\text{C}_ 4\) approximations to bias and mean squared error of a sample mean under nonresponse are obtained.