I understand that the robust variances may lead to a different
standard error. I want the standard error valid for heteroscedastic
data, ultimately, because I have very good estimates of the
measurement variances (why I'm doing weighted fits in the first
place).
For the simple example here, the W
The robust variances are a completely different estimate of standard error. For linear
models the robust variance has been rediscovered many times and so has lots of names: the
White estimate in economics, the Horvitz-Thompson in surveys, working independence
esitmate in GEE models, infinitesi
> Survreg treats weights as case weights, and lm treats them as sampling
> weights.
> Here is a simple example. Data set test2 has two copies of every obs in data
> set test.
>
> > test <- data.frame(x=1:6, y=c(1,3,2,4,6,5))
> > test2 <- test[c(1:6, 1:6),]
>
> > summary(lm( y ~ x, data=test))$co
On Tue, Feb 25, 2014 at 10:50 AM, Therneau, Terry M., Ph.D. <
thern...@mayo.edu> wrote:
>
>
> On 02/25/2014 05:00 AM, r-help-requ...@r-project.org wrote:
>
>> Hi,
>>
>> I have some measurements and their uncertainties. I'm using an
>> uncensored subset of the data for a weighted fit (for now---I'
On 02/25/2014 05:00 AM, r-help-requ...@r-project.org wrote:
Hi,
I have some measurements and their uncertainties. I'm using an
uncensored subset of the data for a weighted fit (for now---I'll do a
fit to the full, censored, dataset when I understand the results).
survreg() reports a much sma
Hi,
I have some measurements and their uncertainties. I'm using an
uncensored subset of the data for a weighted fit (for now---I'll do a
fit to the full, censored, dataset when I understand the results).
survreg() reports a much smaller standard error for the model
parameter than lm(), but only
6 matches
Mail list logo
