00:00
alpha? Usually 0, sometimes 0.01. Now there
is a different way to express statistical
significance without using a P value, we call
this a confidence interval. A confidence interval
gives us a range in which your value probably
has meaning or probably is reality. In science
currently we prefer a confidence interval
to a P value, you never report both of them,
just one or the other. So when you're reading
papers on a variety of epidemiological measures,
you will find either P values or confidence
intervals, never both, and again we prefer
confidence intervals. This is how it works.
A confidence interval contains a point estimate,
that's the thing that we care about, the thing
that we’re trying to measure. For example,
the mean age of university students, 21 years,
that's my computed average age, but it also
gives me a range, an interval under which
X percent of the time I'm going to find an
answer in that range, what X percent? 95%
usually. Where does that 95% come from? Well
that's from my alpha level, remember 0,
it's the same as 95. Don't worry about it
too much, just remember that a confidence
interval gives us a range of values in which
the actual answer probably sits. Now there
is a variety of statistical tests that you
can use or you will come across in your science
career. The T-test is a very famous and common
test for comparing the means of different
groups. Then we have a chi-square test, we
use that to compare whether categorical variables
are related or associated with each other.
01:39
The ANOVA or analysis of variance analysis,
is kind of like a T-test with more than two
groups, we can compare three or more groups
to see if the mean values are the same or
different. And correlations tell us whether
or not continuous variables are related to
one another, is height correlated with age,
is weight correlated with height, is income
correlated with age or lifespan, that sort
of thing. And we have regressions. Regressions
are enormously popular in epidemiology. In
regressions we can determine whether or not
certain variables have an influence on another
important outcome variable.