Normalizing data?

  • Thread starter Thread starter pm
  • Start date Start date
P

pm

Hi everyone. I have a collection of 32 values from an experimental group
(16) and a control group (16). Each set of data was taken over 4 days. So,
for example, Day1 I did experimental 1-4 and control 1-4. Day2, experimental
5-8, control 5-8, etc....

My data collection tool can be calibrated differently on each day
(potentially). I am looking for a way to normalize this data. Any
suggestions? Ideally, if Day 1's control avg was 24 and Day 2's was 27, I'd
like to normalize all of Day 1 and Day 2 based on the control values for that
day.

Thanks in advance....it's been awhile since stats!
 
A lot depends on what your numbers mean, and how your 'instrument' or 'tool' operates - you could
add/subtract the difference (your reading was 2 points higher than expected, so subtract 2 points
from all readings) or you could use a ratio (your readings averages 26 instead of 24, so multiply
all values by 24/26...).

Another approach would be to simply use the numbers as is, and use the variation in your readings
on the control group to set the upper bound on your confidence levels - if your readings vary that
much from day to day, then they could vary somewhat from reading to reading, and can just be treated
as another source of error.

Of course, when I need to do this, I give my data to a statistician to handle... ;-)

HTH,
Bernie
MS Excel MVP
 
I am looking for a way to normalize this data.

Hi. Probably not what you want, but from a math point of view, to
Normalize a vector, you divide each value by the Norm of the Vector.
So, if one had data in A1:A4...

B1 = A1/Sqrt(A1^2 + A2^2 + A3^2 + A4^2)
B2 = A2/Sqrt[A1^2 + A2^2 + A3^2 + A4^2]
ect...

The Norm (denominator) can better be written as

=Sqrt(SUMSQ(A1:A4))

= = = = =
HTH
Dana DeLouis
 
Back
Top