Estimates of sampling errors.

Sampling errors were provided in tables for the entire Turkish sample urban and rural areas regions and age groups for 42 variables included in the 1993 Turkish Demographic and Health Survey (TDHS). Sampling errors were also calculated for the total fertility rate of the last year prior to the survey and the infant mortality rate for the five years preceding the survey nationally and for urban and rural areas; the Jackknife methodology was used to calculate sampling errors. Tables provided the standard error the number of cases weighted and unweighted the design effect relative error and confidence limits. Nonsampling errors due to mistakes in data collection and processing are difficult to evaluate statistically. Sampling errors can be evaluated through the ratio of the standard deviation to the square root of the sample size for a particular statistic: the standard error. Any statistic from a sample survey will fall within a range of plus or minus two times the standard error of that statistic in 95% of all possible samples of the same size and design. The TDHS was a three-stage stratified design. CLUSTERS was the computer software program used to compute the sampling errors for 42 variables from the TDHS. 24 variables were found to have relative errors of less than 0.03 which meant that the standard error was at most 3% of the estimate. 13 variables showed a range of values for the relative error of 0.031 and 0.059. Five variables had a relative error greater than 0.06. The maximum relative error was 16.6% which applied to variables that were very rare events. The design effect value was under 1.3 for 24 variables between 1.31 and 1.5 for 13 variables and over 1.51 for 5 variables. The maximum design effect value was 1.668. The average for the 42 variables was 1.301: 1.213 in urban and 1.293 in rural areas for 41 variables.