Direct standardization within domains

In health surveys it is often of interest to standardize domains to have the same distribution of, eg, age as in a target population. The operation is similar to post-stratification, except that the totals for the domains are fixed at the current estimates, not at known population values. This function matches the estimates produced by the (US) National Center for Health Statistics.

svystandardize(design, by, over, population, excluding.missing = NULL)

Arguments

design

survey design object

by

A one-sided formula specifying the variables whose distribution will be standardised

over

A one-sided formula specifying the domains within which the standardisation will occur, or ~1 to use the whole population.

population

Desired population totals or proportions for the levels of combinations of variables in by

excluding.missing

Optionally, a one-sided formula specifying variables whose missing values should be dropped before calculating the domain totals.

Value

A new survey design object of the same type as the input.

References

National Center for Health Statistics https://www.cdc.gov/nchs/tutorials/NHANES/NHANESAnalyses/agestandardization/age_standardization_intro.htm

Note

The standard error estimates do not exactly match the NCHS estimates

See also

postStratify, svyby

Examples

## matches http://www.cdc.gov/nchs/data/databriefs/db92_fig1.png
data(nhanes)
popage <- c( 55901 , 77670 , 72816 , 45364 )
design<-svydesign(id=~SDMVPSU, strata=~SDMVSTRA, weights=~WTMEC2YR, data=nhanes, nest=TRUE)
stdes<-svystandardize(design, by=~agecat, over=~race+RIAGENDR, 
   population=popage, excluding.missing=~HI_CHOL)
svyby(~HI_CHOL, ~race+RIAGENDR, svymean, design=subset(stdes,
agecat!="(0,19]"))
#>     race RIAGENDR   HI_CHOL          se
#> 1.1    1        1 0.1543786 0.008318204
#> 2.1    2        1 0.1142946 0.010182838
#> 3.1    3        1 0.1020776 0.013547678
#> 4.1    4        1 0.1358312 0.042274271
#> 1.2    1        2 0.1316436 0.013418637
#> 2.2    2        2 0.1543247 0.008932134
#> 3.2    3        2 0.1025411 0.018953586
#> 4.2    4        2 0.1197434 0.040091106


data(nhanes)
nhanes_design <- svydesign(ids = ~ SDMVPSU, strata = ~ SDMVSTRA, 
                        weights = ~ WTMEC2YR, nest = TRUE, data = nhanes)

## These are the same
nhanes_adj <- svystandardize(update(nhanes_design, all_adults = "1"),
                 by = ~ agecat, over = ~ all_adults,
                 population = c(55901, 77670, 72816, 45364), 
                 excluding.missing = ~ HI_CHOL)
svymean(~I(HI_CHOL == 1), nhanes_adj, na.rm = TRUE)     
#>                         mean     SE
#> I(HI_CHOL == 1)FALSE 0.89413 0.0053
#> I(HI_CHOL == 1)TRUE  0.10587 0.0053

nhanes_adj <- svystandardize(nhanes_design,
                 by = ~ agecat, over = ~ 1,
                 population = c(55901, 77670, 72816, 45364), 
                 excluding.missing = ~ HI_CHOL)
svymean(~I(HI_CHOL == 1), nhanes_adj, na.rm = TRUE)     
#>                         mean     SE
#> I(HI_CHOL == 1)FALSE 0.89413 0.0053
#> I(HI_CHOL == 1)TRUE  0.10587 0.0053