Ali wrote (after a suggestion to use -stcox- to study determinants of
gestational length (duration of pregnancies)):
That is true, specially with the fact that the exposure is during pregnancy so
it depends on gestational length. However, to avoid this I am looking at
exposures during the first 23 weeks' gestation were all women have the same
opportunity to be exposed. when I want to see the effect of exposure on
prematurity (binary variable I used binomial regression instead of cox
regression by restricting the exposure to 23 weeks). In addition I want to see
the effect of exposure on gestational age (continuous). This calls for linear
regression-as I understand- but as I mentioned the residuals distribution is
not normal. In these cases a non-parametric analysis is required. I used qreg
but the results do not look right.
Since the dataset is very large does it matter if the residuals normal
distribution assumption was violated in linear regression?
--------------------------------------------------------------------
There is no doubt that the residuals after a linear regression are far from
a normal distribution. Also, there is hardly any meaningful transformation of
gestational age that could make this happen. So, relying on linear regression
is a problem.
-stci- can be used for simpler descriptions and comparisons, using e.g. the
median and other percentiles in a sort of non-parametric approach.
Good luck,
Svend
________________________________________________________
Svend Juul
Institut for Folkesundhed, Afdeling for Epidemiologi
(Institute of Public Health, Department of Epidemiology)
Vennelyst Boulevard 6
DK-8000 Aarhus C, Denmark
Phone, work: +45 8942 6090
Phone, home: +45 8693 7796
Fax: +45 8613 1580
E-mail: [email protected]
_________________________________________________________
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
