View source: R/reweight_dual.R

reweight | R Documentation |

Evaluates a set of expressions for different frame weights in a dual-frame/multi-frame design, so that an optimal or compromise-optimal set of frame weights can be chosen

```
reweight(design, ...)
## S3 method for class 'dualframe'
reweight(design, targets=NULL, totals=NULL,
estimator=c("constant","expected"),
theta=NULL, theta_grid=seq(0,1,by=0.05),...)
## S3 method for class 'dualframe_with_rewt'
plot(x,y,type="b",...)
```

`design` |
dual-frame or multiframe design object |

`targets` , `totals` |
A list of quoted expressions estimating the
variance of a survey estimator ( |

`estimator` |
As in |

`theta` |
As in |

`theta_grid` |
Grid for optimising theta over, with |

`x` |
object produced by |

`y` |
ignored |

`type` , `...` |
in the |

Traditionally, this optimisation has been done with totals, which is a good default and more mathematically tractable. However, when the point of multiple-frame sampling is to improve precision for a rare sub-population, or when you're doing regression modelling, you might want to optimise for something else.

An object of class "dualframe_with_rewt".

The `coef`

method returns the optimal theta for each target.
The `rewt`

element includes the variances of each target on a grid of
`theta`

in `variances`

`multiframe`

```
data(phoneframes)
A_in_frames<-cbind(1, DatA$Domain=="ab")
B_in_frames<-cbind(DatB$Domain=="ba",1)
Bdes_pps<-svydesign(id=~1, fpc=~ProbB, data=DatB,pps=ppsmat(PiklB))
Ades_pps <-svydesign(id=~1, fpc=~ProbA,data=DatA,pps=ppsmat(PiklA))
## Not very good weighting
mf_pps<-multiframe(list(Ades_pps,Bdes_pps),list(A_in_frames,B_in_frames),theta=0.5)
svytotal(~Lei+Feed+Tax+Clo,mf_pps, na.rm=TRUE)
## try to optimise
mf_opt<-reweight(mf_pps, totals=list(~Lei, ~Feed,~Tax,~Clo))
coef(mf_opt)
plot(mf_opt)
## a good compromise is 0.80 for everything except Tax
## and it's still pretty good there
## (Tax will be biased because it's missing for landline-only)
mf_pps_opt<-reweight(mf_opt,theta=0.80)
svytotal(~Lei+Feed+Tax+Clo,mf_pps_opt, na.rm=TRUE)
## Targets other than totals
mf_reg<-reweight(mf_pps,
targets=list(quote(vcov(svyglm(Lei~Feed+Clo, design=.DESIGN))[1,1]),
quote(vcov(svytotal(~Lei,.DESIGN))))
)
plot(mf_reg,type="l")
legend("topright",bty="n",lty=1:2,col=1:2, legend=c("regression","total"))
## Zooming in on optimality for a particular variable (for compatibility)
mf_opt1<-reweight(mf_pps, totals=list(~Feed),theta_grid=seq(0.7,0.9,length=100))
coef(mf_opt1) # Frames2::Hartley gives 0.802776
```

Embedding an R snippet on your website

Add the following code to your website.

For more information on customizing the embed code, read Embedding Snippets.