3stage2phase.R: twophase designs with three-stage sample at phase 1

api.R: Run example(api) to check that results haven't changed

bycovmat.R: Check that svyby(,covmat=TRUE) is getting the ordering
   of estimates correct.

caleg.R: Calibration examples
	 - calibration to information on PSUs rather than population
	 - check that bounded weights really are bounded
	 - check that linear calibration with error proportional to
           x agrees with ratio estimators

check.R:  Many combinations of options for svydesign

deff.R:  Regression test on design effects, especially for totals

DBIcheck.R: Check that we get the same results for SQLite-backed and
         in-memory versions of the API data.

domain.R: Check that domain estimators of means and their standard
          errors agree with derivations as ratio and regression
          estimators.  Repeat for calibrated and raked designs

fpc.R: Many ways to specify fpc

kalton.R: Calibration examples from Kalton & Flore-Cervantes,
          J Off Stat, 19(2) 81-97

lonely.psu.R: All the lonely PSU options

multistage.R: Check that a two-stage cluster sample analysis agrees with
              the hand-calcuated result in Sarndal et al.

nwts.R:	Compare results from twophase() to published two-phase
        case-control example

nwts-cch.R: Compare results from twophase() to case-cohort analyses in
        survival package.

pps.R: Brewer's approximation for pps without replacement

quantile.R: quantile estimation on a lognormal sample

rakecheck.R: check that raking by iterative post-stratification agrees with
 	     raking using calibrate()

regpredict.R: ratio and regression estimation of a total.

scoping.R: check that svyglm and svycoxph work inside functions.

survcurve.R: check that svykm and predict.coxph give the same result
           when a data set is doubled and the two replicates of each
           observation are treated as a cluster.

twophase.R: separately verifiable examples of twophase studies

quantile-chile.R: example of interpolating for quantiles, from Chilean
	stats agency

na_action.R: check that svyglm works with na.action=na.exclude