Re: st: Plot means with SD

[Patricia Biedermann <pati.biedermann@gmail.com>]

Dear Hoaglin,

Thank you for your reply and ideas.
This makes all sense to me...and I also would rather go for median instead of mean...above all if the data is skewed. But unfortunately I'm the last link in the chain...

Partricia

On 04.04.2012, at 17:35, David Hoaglin wrote:

> Dear Patricia,
> 
> Thank you for the explanation.
> 
> I understand that the decision to present the results as mean and SD
> is not yours.  If, however, the display will show mean and SD for
> medstable and, separately, mean and SD for durofstay, I find it
> difficult to see how such a display would be appropriate without
> further analysis.  The main problem is that it would ignore the
> connection between those variables' values for the same patients.  The
> differences within the individual patients are often the most
> important part of the difference between the variables, so one needs a
> strong justification for not looking at them.  To see the "whole
> effect," you need to clear away the variation among patients.  Another
> shortcoming is that the SD isn't a very good summary of variation when
> the data are substantially skewed.  For some analyses it is helpful to
> transform such variables, by using the logarithm or perhaps the
> reciprocal.
> 
> I'm not sure how one determines "that people stay longer in hospital
> as actually needed," but a scatterplot is likely to be important.  A
> plot of medstable versus (durofstay MINUS medstable) would show
> whether medstable is associated with the length of stay before the
> patient achieves medical stability.
> 
> David Hoaglin
> 
> On Wed, Apr 4, 2012 at 7:12 AM, Patricia Biedermann
> <pati.biedermann@gmail.com> wrote:
>> Dear David,
>> 
>> Both variables are positive (right) skewed) and the variables come from the same patients. We want to see the difference between the two variables (that people stay longer in hospital as actually needed). The difference of each person is not so important now....we want to see a whole effect.
>> 
>> The results should be presented in mean and SD...(was not my decision...I just have to do the graphs)
> 
> *
> *   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/


*
*   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/