R help - Apr 2011 - Extrapolating data points for individuals who lack them

Hi,

We have an experiment where individuals responses were measured over 5 days.
Some responses were not obtained because we only allowed individuals to respond
within a limited time-frame. These individuals are given the maximum response
time as they did not respond, yet we feel they may have done if given time (and
by looking at the rest of their responses over time, the non-response days stand
out).

We therefore want to extrapolate data points for individuals, on days when they
didn't respond, using a regression of days when they did.

Does anyone know how we could do this quickly and easily in R?

Thanks very much
Dave

	[[alternative HTML version deleted]]

----------------------------------------
> From: de223 at exeter.ac.uk
> To: r-help at r-project.org
> Date: Wed, 20 Apr 2011 12:41:29 +0100
> Subject: [R] Extrapolating data points for individuals who lack them
>
> Hi,
>
> We have an experiment where individuals responses were measured over 5
days. Some responses were not obtained because we only allowed individuals to
respond within a limited time-frame. These individuals are given the maximum
response time as they did not respond, yet we feel they may have done if given
time (and by looking at the rest of their responses over time, the non-response
days stand out).
>
> We therefore want to extrapolate data points for individuals, on days when
they didn't respond, using a regression of days when they did.
>
> Does anyone know how we could do this quickly and easily in R?
You are probably talking about right censoring. See things like this, 
( you may have good luck just with "R" rather than "CRAN" ) 

http://www.google.com/#sclient=psy&hl=en&source=hp&q=CRAN+informative+%22right+censoring%22


If you post data maybe someone can try a few things. It isn't hard to take
data
subsets, fit models, and replace data with model predictions but easier and more
interesting to illustrate with your data.


Personally I would avoid making up data and of course extrapolation tends
to be the most error prone way of doing that. 




>
> Thanks very much
> Dave