00:00
I'm seeking to reject that null hypothesis.
Now imagine you're doing a test, a measurement,
maybe comparing a treatment group versus a
placebo group in your randomized controlled
trial, and you get a result. And let's say
you do that test again, you get a different
result. You did that test 1 million billion
number of times, you get 1 million billion
different results. Sometimes you get the same
result, sometimes you don't. If you're to
plot the frequency with which certain results
occur, you'll get a shape that looks like
a bell curve or a normal curve. Now the P
value is a value under that normal curve that
tells you whether or not it's probable that
the null hypothesis is true. More precisely,
it's telling you the probability that you
rejecting that null hypothesis was done incorrectly.
00:50
It’s a difficult concept to absorb but just
rely upon this bit of rule; if your P value
is less than a certain cut-off value, you can
reliably reject the null hypothesis. What's
that cut-off value? We call that value an alpha
value and it is usually set at 0.05 or 0.01.
01:12
There is no the reason for that, that is just
history, 0.05 is the most common one. So again,
if your P value is less than 0.05, you can
reliably reject the null hypothesis and conclude
that you probably found something, you probably
found an effect. For example, if we're testing
whether the average heights of two different
groups of children are different and we perform
a test to do so, let’s say a T-test, which
is the appropriate test in this case, and
you find a P value of 0.02. 0.02 is less than
0, so we conclude that we can probably reject
our null hypothesis. How likely is it that
that rejection was done in error? Well that’s
the P value, 0, 2%, there is a 2% chance
that I rejected that null hypothesis in error,
we say that's a good enough number. Now defining
or interpreting a P value is problematic at
the best of times, there is a convenient,
though inaccurate interpretation that the
P value is the probability that your result
was due to chance alone. So in the previous
example, there is a 2% chance that your result
was all luck, that's not really what's going
on here. More accurately the P value is a
probability that your test incorrectly rejected
the null, that's all it’s saying. Now if
that's too confusing, just remember this very
simple rule, if the P is low, the null hypothesis
must go, you want a low P value, it means
you found something of statistical importance.
What is low? Low is a P less than alpha. What’s
alpha? Usually 0.05, sometimes 0.01. Now there
is a different way to express statistical