Mail survey is nowadays frequently used in collecting data in
the field of social scientific study. But the greatest defect
of mail survey is its low response rate, often with a
percentage ranging roughly from twenty to forty percent. And
because of the low response rate, there are many unreturned
questionnaires that will cause nonrseponse bias which is going
to affect both the representativity of the samples and the
reliability of the deducing result. To heighten the response
rate of mail survey can lower the influence of nonresponse
bias. In fact, it is unlikely to track down each individual who
does not return their questionnaires by mail, when it is put
into practice. Therefore, based on Hansen and Hurwitz''s idea
about partial trailing, this article provides two ways of
dealing with incompletely sample data in mail survey. That is
(1) To employ fixed rate , draw a traced sample size;(2) To
draw a fixed amount of tracing sample size, and collect the
data about those who do not return their questionnaires by
either telephone or personal interview. Meanwhile, with these
two and under a sensible probability distribution of the
returned sample size in mail survey, use the simple random
sampling and stratified random sampling to estimate both
population mean and population proportion. Under the conditions
of estimate accuracy, demanded by survey cost, the requirements
for sampling design, the tolerable sampling error, decide the
number of suitable size in mail survey and the fixed number of
the samples going to be trailed in order to apply to the
practice of doing survey. Finally, this thesis investigates the
distribution of an actual sample size increment in stratified
random sampling, and it''s expected value, and variance.