Bangladesh 1993-1994 Demographic and Health ... - Measure DHS
Bangladesh 1993-1994 Demographic and Health ... - Measure DHS
Bangladesh 1993-1994 Demographic and Health ... - Measure DHS
- No tags were found...
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
APPENDIX BESTIMATES OF SAMPLING ERRORSThe estimates from a sample survey are affected by two types of errors: nonsampling errors <strong>and</strong>sampling errors. Nonsampling errors are the results of mistakes made in implementing data collection <strong>and</strong>data processing, such as failure to locate <strong>and</strong> interview the correct household, misunderst<strong>and</strong>ing of thequestions on the part of either the interviewer or the respondent, <strong>and</strong> data entry errors. Although numerousefforts were made during the implementation of the B<strong>DHS</strong> to minimize this type of error, nonsampling errorsare impossible to avoid <strong>and</strong> difficult to evaluate statistically.Sampling errors, on the other h<strong>and</strong>, can be evaluated statistically. The sample of respondents selectedin the B<strong>DHS</strong> is only one of many samples that could have been selected from the same population, using thesame design <strong>and</strong> expected size. Each of these samples would yield results that differ somewhat from theresults of the actual sample selected. Sampling errors are a measure of the variability between all possiblesamples. Although the degree of variability is not known exactly, it can be estimated from the survey results.Sampling error is usually measured in terms of the st<strong>and</strong>ard error for a particular statistic (mean,percentage, etc.), which is the square root of the variance. The st<strong>and</strong>ard error can be used to calculateconfidence intervals within which the true value for the population can reasonably be assumed to fall. Forexample, for any given statistic calculated from a sample survey, the value of that statistic will fall within arange of plus or minus two times the st<strong>and</strong>ard error of that statistic in 95 percent of all possible samples ofidentical size <strong>and</strong> design.If the sample of respondents had been selected as a simple r<strong>and</strong>om sample, it would have beenpossible to use straightforward formulas for calculating sampling errors. However, the B<strong>DHS</strong> sample is theresult of a two-stage stratified design, <strong>and</strong>, consequently, it was necessary to use more complex formulas.The computer software used to calculate sampling errors for the B<strong>DHS</strong> is the ISSA Sampling Error Module(ISSAS). This module used the Taylor linearization method of variance estimation for survey estimates thatare means or proportions. The Jacknife repeated replication method is used for variance estimation of morecomplex statistics such as fertility <strong>and</strong> mortality rates.The Taylor linearization method treats any percentage or average as a ratio estimate, r = y/x, wherey represents the total sample value for variable y, <strong>and</strong> x represents the total number of cases in the group orsubgroup under consideration. The variance of r is computed using the formula given below, with thest<strong>and</strong>ard error being the square root of the variance:in whichvar,r, h I'; )]= Y a-r'xh , <strong>and</strong> z h = Yh-r'x hwhere hmhy~.frepresents the stratum which varies from 1 to H,is the total number of enumeration areas selected in the hthstratum,is the sum of the values of variable y in EA i in the h th stratum,is the sum of the number of cases in EA i in the h th stratum, <strong>and</strong>is the overall sampling fraction, which is so small that it is ignored.151