Re: FW: st: Plotting several linear associations in one graph for comparison

[Maarten Buis <maartenlbuis@gmail.com>]

On Mon, Jul 16, 2012 at 4:10 PM, Amal Khanolkar wrote:
> - I understand the need for the replace commands. But as above, I think I introduce bias by narrowing the study population down to a very small proportion of the true study population. I tried choosing for example, mean BMI, and height, and the largest category of maternal education with most number of subjects. Most of my subjects are aged 18 (around 90% of them), but restricting it to just one birth year, means I loose most of the population. Is it possible in my example above to select select a range of birth years (all birth years for example, from 1973 to 2008), maybe include more maternal educational categories?? I could then obtain a graph for a study population with greater variation right?

Bias means that you do not measure what you want to measure. As a
consequence, no estimate can be biased per se, i.e. without specifying
what the estimate should be an estimate of. In this case I do not
think your predictions are biased (assuming your model is ok): they
measure the predicted blood pressure for the specific group you chose.

The problem you are facing is that with 8 covariates your regression
model is a plane in 9 dimensional space, which is virtually impossible
to visualize in one go. (I have heard certain mushrooms can help, but
I have not tried it myself and I would not recommend their use.) The
trick is not to try to visualize the entire model in one go, but to
look at different "slices". You have looked at BP by motherland and
birth weight, you can now create a different graph that fixes birth
weight and varies birth year, and another graph that varies BMI, etc..

-- Maarten

--------------------------
Maarten L. Buis
Institut fuer Soziologie
Universitaet Tuebingen
Wilhelmstrasse 36
72074 Tuebingen
Germany


http://www.maartenbuis.nl
--------------------------

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/