Estimates of Sampling Error
Sampling errors are presented in Tables B.02 - B.16 for variables considered to be of major interest. Results are presented for the whole country, for urban and rural areas separately, for each of four education groups, for each of five regions, and for each of three age groups. For each variable, the type of statistic (percentage, mean or rate) and the base population are given in Table B.01. For each variable, Tables B.02 - B.16 present the value of the statistic (R), its standard error (SE), the number of cases (N) where relevant, the design effect (DEFT) where applicable, the relative standard error (SE/R), and the 95 percent confidence limits (R-2SE, R+2SE).
The confidence limits have the following interpretation. For the percentage of currently married women using the contraceptive intrauterine device (IUD), the overall value for the full sample is 32.8%, and its standard error is 0.7%. Therefore, to obtain the 95 percent confidence limits, one adds and subtracts twice the standard error to the sample estimate, which means that there is a high probability (95 percent) the true percentage currently using the IUD is between 31.3% and 34.2%.
The relative standard errors for most estimates for the country as a whole are small, except for estimates of very small percentages. The magnitude of the error increases as estimates for sub-populations such as geographical areas are considered. For the variable IUD, for instance, the relative standard error (as a percentage of the estimated parameter) for the whole country and for urban and rural areas is 2.2 percent, 3.2 percent, and 3.0 percent, respectively. For the five regions, the relative standard error of the variable IUD varies between 3.9 percent and 7.4 percent.
Special mention should be made of the sampling errors for rates. The denominators are exposure-years, and the numerators are either births or deaths in the population under consideration during the indicated period of time. Estimates of sampling errors are shown for the TFR in the three years prior to the survey, presented in Chapter 3, and for the various 3-year mortality rates presented in Chapter 7. These estimates are calculated at the national level, and by urban-rural residence, region, and mother's educational level. (They are irrelevant for age groups)
It should be noted that the survey indicates, with a 95 percent level of confidence, that the TFR for the 3-year period prior to the survey lay between 2,4 and 2,6 children per woman, and that the infant mortality rate for the 3-year period prior to the survey lay between 23 and 35 per thousand births. The differences between the survey results and registration statistics are not due to sampling variability.
Details of the sampling errors are presented in the appendix B to the national report RHS 2003 of the external resources.
|
