Compute an appropriately scaled empirical variance estimate from
replicates. The `mse`

argument specifies whether the sums of
squares should be centered at the point estimate (`mse=TRUE`

) or
the mean of the replicates. It is usually taken from the `mse`

component of the design object.

```
svrVar(thetas, scale, rscales, na.action=getOption("na.action"),
mse=getOption("survey.replicates.mse"),coef)
```

## Arguments

- thetas
matrix whose rows are replicates (or a vector of replicates)

- scale
Overall scaling factor

- rscales
Scaling factor for each squared deviation

- na.action
How to handle replicates where the statistic could
not be estimated

- mse
if `TRUE`

, center at the point estimated, if
`FALSE`

center at the mean of the replicates

- coef
The point estimate, required only if `mse==TRUE`