|
|
"Tom Claessens" <tom.claesDELETETHISsens@xxxxxxxx> wrote in message
news:dcgde2$rep$1@xxxxxxxxxxxxxxxxxx
> Hi,
>
> I have a question about linear regression.
>
> I want to regress y (for example blood pressure) on x1 (length) and x2
> (weight).
> Suppose that, for some reason, the x1 data is truncated, i.e. every
> person that is taller than 2 meters is said to be exactly two meters.
> Everyone smaller than two meters keeps his length.
>
> Do I have to take this into account when performing linear regression,
> and if so, how?
>
> Thank you very much in advance,
> Tom
Yes is does matter.
The first step is to ignore the tall people.
I would try and use this data to determine if blood pressure depends on
height over and above the relationship between height and weight.
I.e does blood pressure depend on being overweight alone or not.
How you go from there depends on the answer you get, but I think you can
tease a bit more information out of the data.
For example you can figure out if some of the tall people are underweight
and use this in a modified regression.
|
|