epi.stratasize {epiR} | R Documentation |
Estimates the required sample size under under stratified random sampling.
epi.stratasize(strata.n, strata.mean, strata.var, epsilon, method = "mean", conf.level = 0.95)
strata.n |
vector, defining the size of each strata. |
strata.mean |
vector, representing the mean of each strata. |
strata.var |
vector, representing the variance of each strata. |
epsilon |
the maximum relative difference between our estimate and the unknown population value. |
method |
a character string indicating the method to be used. Options are mean , total , proportion , or pps . |
conf.level |
scalar, defining the level of confidence in the computed result. |
A list containing the following:
strata.sample |
the estimated sample size for each strata. |
strata.total |
the estimated total size. |
strata.stats |
mean the mean across all strata, sigma.bx the among-strata variance, sigma.wx the within-strata variance, and sigma.x the among-strata variance plus the within-strata variance, rel.var the within-strata variance divided by the square of the mean, and gamma the ratio of among-strata variance to within-strata variance. |
Use method proportion
to estimate sample size using stratified random sampling with equal weights (see Levy and Lemeshow, page 176). Use method pps
to estimate sample size using proportional stratified random sampling with proportional allocation (see Levy and Lemeshow, page 179).
Where method = "proportion"
the estimated proportions for each strata are entered into the vector strata.mean
. The vector strata.var
is ignored.
Mark Stevenson (EpiCentre, IVABS, Massey University, Palmerston North, New Zealand)
Javier Sanchez (Atlantic Veterinary College, University of Prince Edward Island, Charlottetown Prince Edward Island, C1A 4P3, Canada).
Levy PS, Lemeshow S (1999). Sampling of Populations Methods and Applications. Wiley Series in Probability and Statistics, London, pp. 175 - 179.
## EXAMPLE 1 ## Hospital episodes (Levy and Lemeshow 1999, page 176 -- 178) ## We plan to take a sample of the members of a health maintenance ## organisation (HMO) for purposes of estimating the average number ## of hospital episodes per person per year. The sample will be selected ## from membership lists according to age (under 45 years, 45 -- 64 years, ## 65 years and over). The number of members in each strata are 600, 500, ## and 400 (respectively). Previous data estimates the mean number of ## hospital episodes per year for each strata as 0.164, 0.166, and 0.236 ## (respectively). The variance of these estimates are 0.245, 0.296, and ## 0.436 (respectively). How many from each strata should be sampled to be ## 95% that the sample estimate of hospital episodes is within 20% of the ## true value? strata.n <- c(600, 500, 400) strata.mean <- c(0.164, 0.166, 0.236) strata.var <- c(0.245, 0.296, 0.436) epi.stratasize(strata.n, strata.mean, strata.var, epsilon = 0.20, method = "mean", conf.level = 0.95) ## The number allocated to the under 45 years, 45 -- 64 years, and 65 years ## and over stratums should be 223, 186, and 149 (a total of 558). These ## results differ from the worked example provided in Levy and Lemeshow where ## certainty is set to approximately 99%. ## EXAMPLE 2 ## Dairies are to be sampled to determine the proportion of herd managers ## using foot bathes. Herds are stratified according to size (small, medium, ## and large). The number of herds in each strata are 1500, 2500, and 4000 ## (respectively). A review of the literature indicates that use of foot bathes ## on farms is in the order of 0.50, with the probability of usage increasing ## as herds get larger. How many dairies should be sampled? strata.n <- c(1500, 2500, 4000) strata.mean <- c(0.50, 0.60, 0.70) epi.stratasize(strata.n, strata.mean, epsilon = 0.20, method = "proportion", conf.level = 0.95) ## A total of 54 herds should be sampled: 10 small, 17 medium, and 27 large.