Statalist The Stata Listserver


[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

RE: st: RE: RE: RE: Multiple comparisons, non-normal data


From   "Nick Cox" <n.j.cox@durham.ac.uk>
To   <statalist@hsphsun2.harvard.edu>
Subject   RE: st: RE: RE: RE: Multiple comparisons, non-normal data
Date   Fri, 25 May 2007 12:07:43 +0100

I think you need to talk to someone locally, not 
to try to work out a statistical strategy for a substantial 
project through a discussion list. I know that many people
ask how to analyse their data or what they should be doing
next in their project through Statalist, but such use of
the list often makes me very uneasy (personal opinion). 

If your advisor, supervisor or advisory committee does not 
include anybody with statistical expertise, then they 
should be pestered to suggest someone who has such 
expertise. 

I can identify three important considerations, but 
I won't hazard a guess at precisely what you should be doing: 

1. The analysis should be appropriate for the experimental
design. So, either you designed this or someone else did, 
and if the project was well planned, planning would have
included some consideration of the intended analysis. 

2. If as you imply the data are (seriously) non-normal, 
that usually imply transformation of the data and/or
use of a non-identity link function. 

3. You appear to want to take different times as independent. 
That is a very strong assumption and on the face of it 
implausible. 

Nick 
n.j.cox@durham.ac.uk 

Knag Anne-Christine
 
> Haha, you are quite right! I am a master student at the 
> Institute of Marine
> Research in Norway. My thesis is on the effect of oil pollution on cod
> (which I guess you all know is a species of fish). I am 
> really struggling
> with my stats as everyone is telling me something different. 
> The latest tip
> was to use a Kolmogorov Smirnov. IŽll so rude and post my 
> question again:
> 
> I have data on an experiment where 9 tanks were sampled every 
> day for 3
> months. I measured different variables every day. Some of 
> them should be
> looked at as how much variable x at day n contributed to the 
> sum of tank.
> (ie relative fecundity)
>  
> Each of the nine tanks contains one group of individuals that 
> have been
> given either treatment C, L, M, H or E at time 1 or 2. I.e 
> tank 9 is a group
> given treatment E at time 2, while tank 4 is a group given E 
> at time 1.
> There are no replicates of the treatment, but if there is no 
> effect of time
> (1 or 2) given the treatment, I believe that I can use the 
> tanks with the
> same treatment given at different times se as replicates of treatment.
>  
> In addition to have measured the fecundity (number of eggs 
> produced), I also
> have the diameter of the eggs collected. I would like to know how the
> diameter changes over time in each of the tanks and whether it differs
> between the tanks.
> 
> I would also like to check for significant differences 
> between the nine
> tanks in the day they initiated spawning, the total number of days of
> spawning and when spawning ended.
>  
> Again, I will be very, very grateful for any help that you 
> can give me.

*
*   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/



© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index