Significance of Findings
Name: Tierra H.
If a finding is statistically significant, why is it also necessary to
consider practical significance?
One has to be aware that statistics is a discipline for handling our ignorance. If we
know that A causes B, or A and B are correlated (that is, if A changes by so much, then
B changes by some amount), we do not need statistics.
The language of statistics, especially regarding cause and effect, is very precise, very
careful, and a bit daunting.
For example: A fictional statistician might say something like, "If repeated measurement
of some variable, X, is normally and randomly distributed about the mean, M, with
standard deviation, S, then with a confidence level of 0.YZ (usually 0.95 or 0.99)
we can say that any values of X that differ from the mean, M, may be explained by
random fluctuations in the variable, X, and
hence are not statistically significant.
Now there are a lot of "if" conditions in such a statement: Differences
are normally and randomly distributed about M means there are no more than
one "hump" in the measurements, that the difference between X and M
MAY be due to random fluctuations alone, and some more hidden "ifs". There
are also other types of distributions that do not have a "bell-shaped" curve
like the above. What statistics does is provide us with a "warning" that
something MAY be happening with the variable X that makes its value
"significantly" different than M. It's not a certainty. If X differs from M
more than what random fluctuations can "explain" then it says, "Hey, you
better check this out!" You do not have to, it is just a consistent way of
identifying when some X is significantly different than the average, M.
Click here to return to the Mathematics Archives
Update: June 2012