A Study on the Non-response in Two Characters in a Two Stage Design

Abstract

In sample survey practice, non-response from some of the sampling units is of common occurrence. This problem of ‘non-response’ or ‘Incomplete samples’ is more common in Mailing Enquires. If the non-response is considerable the estimates obtained on the basis of the realized sample may not represent the characteristics of the entire population. The only way to minimize or eliminate bias is to have ‘call-acks’ on then non-responding units, one or more than one, it need arises.

Various attempts have been made to develop suitable plan of enquiry, which will take both bias and the cost of the survey into consideration, whenever a problem of ‘non-response’ arises. Most of the works that has been done in this direction deals with the singly stage sampling and one character under study. The present thesis deals with the problem of non-response in two stage sampling and with two characters under study. The procedure consists in classifying the first sample of PSU’s into 15 strata on the basis of information supplied by SSU’s. The final sample of SSU’s is again classified into four classes according to the information supplied by them. An estimate for the population mean has been built up. The expressions for variance and estimate of variance has been derived. A combined cost function has been developed, To determine the optimum sixes of samples drawn from the different classes.