[BiO BB] Mean and SD
wusu at ele.uri.edu
wusu at ele.uri.edu
Mon Feb 27 12:55:51 EST 2006
Dear Subhash and other interested members:
I am sorry for a late reply. I did type up my reply last Thursday, but the
network went dead as I sent it. I just came back from a conference so let me
try to answer the question again. Thank you for your patience!
I am a biomedical engineer that also teaches undergraduate statistics, so I get
to know some statistics. In order to trim down the discarded portion of the
data set, you may decrease the class width from .02 to .01 and then abandon the
end classes. Since we don't have your original data set, it's hard to tell from
a frequency histogram.
Thanks to the beautiful Central Limit Theorem
so we know that the sampling (or probability) distribution of the sample means
becomes normal if the sample size grows large.
Your sample size was 394 so it's plenty large. Your 394 data values produces
one sample mean, which is only one value of many 'possible' sample means if you
repeat the measurements 394 times over, over and over again. For example, see
http://www.statsoft.com/textbook/stnonpar.html#when
That is to say if you repeat the same experiment many times, the probability
distribution of all the sample means will look normal. Since we don't do such
stupid things, we can really appreciate the word, 'possible' from random
sampling.
Redesign experiments can mean to treat physical boundary problems carefully so
you won't have to throw so many values away. It also applies to simulations, in
which boundary conditions have to be met. I had to do something similar for a
bioelectromagnetics project. Since I don't know what this data set is about, I
can't say much about it. In statistics, experimental design or DOE can mean
different things. You can take a look at a statictics book for
reference or see
http://www.itl.nist.gov/div898/handbook/pri/pri.htm
Perhaps some statistician may respond to us free of charge.
Best regards,
Su Wu - her 2 cents. :)
Quoting Subhash Agarwal <smagarwal at yahoo.com>:
> Dear Forum Members
>
> I agree with the points suggested by all you. From Ryan I would like to ask
> that keeping in mind that the data doesnt follow normal distribution what
> other statistical parameters can be calculated to understand whether the data
> is meaningful or not.
>
> Secondly can Su explain a bit more regarding his suggestion on "redesign
> the experiment or simulation in my opinion".
>
> thanks
> Subhash
>
>
> ---------------------------------
> Jiyo cricket on Yahoo! India cricket
> Yahoo! Messenger Mobile Stay in touch with your buddies all the time.
More information about the BBB
mailing list