"[log in to unmask]" has recently sent several
queries to the list. To me, these increasingly
look like bits of homework, which would be
inappropriate use of the list.
If I'm wrong, and the queries are in fact requests for
help with real-life problems, perhaps the enquirer
could reassure list members by revealing his or her
identity, and giving more details of the actual
conditions (e.g. what condition the doctors are
treating), which could help list members give
informed answers.
Eryl Bassett
---------------------------- Original Message ----------------------------
Subject: more methodology
From: "-" <[log in to unmask]>
Date: Fri, November 6, 2009 12:35
To: "[log in to unmask]" <[log in to unmask]>
--------------------------------------------------------------------------
hello all,
I would like some advice on a variation of a problem I had previously
enquired about:
Someone has data on 500 patients doctor A has been treating and 500
patients doctor B has been treating. There are scores on the patients
regarding symptoms, complications etc and we want to assess the
performance of doctor A from his 1st patient to the last, i.e. for
operative time we want to know if it has increased or not during the
time he has been seeing patients. I received some great advice from
you on linear regression, serial correlation etc.
My new questions are:
- If one randomly assigns the 500 patients of doctor A in 5 groups of
100 patients each, and wants to compare e.g. operative time between
these 5 groups, but if it is also necessary to control for variables
like age, would the appropriate method be ANOVA?
- If one wants to compare a group of doctor A with a group of doctor
B, I guess one should apply ANOVA here as well?
- And if we would like to compare the total performances of the two
doctors for a particular score, e.g. operative time, what would be an
appropriate method? Is it ANOVA again?
Thank you all
|