226 PART 5 Looking for Relationships with Correlation and Regression
When reporting regression coefficients in professional publications, you may
state the SE like this: “The predicted increase in systolic blood pressure with
weight (±1 SE) was 0 49
0 18
.
.
mmHg/kg.”
If you know the value of the SE, you can easily calculate a confidence interval (CI)
around the estimate (see Chapter 10 for more information on CIs). These expres-
sions provide a very good approximation of the 95 percent confidence limits
(abbreviated CL), which mark the low and high ends of the CI around a regression
coefficient:
Lower
CL
Coefficient
SE
%
95
2
Upper
CL
Coefficient
SE
95
2
%
More informally, these are written as 95%
coefficient 2
CI
SE.
So, the 95 percent CI around the slope in our example is calculated as
0 49
2
0 176
.
.
, which works out to 0 49
0 35
.
.
, with the final confidence limits
of 0.14 to 0.84 mmHg. If you submit a manuscript for publication, you may express
the precision of the results in terms of CIs instead of SEs, like this: “The predicted
increase in SBP as a function of body weight was 0.49 mmHg/kg
95% CI : 0.14 0.84 .”
The Student t value
In most output, there is a column in the regression table that shows the ratio of
the coefficient divided by its SE. This column is labeled t value in Figure 16-4, but
it can be labeled t or other names. This column is not very useful. You can think of
this column as an intermediate quantity in the calculation of what you’re really
interested in, which is the p value for the coefficient.
The p value
A column in the regression tables (usually the last one) contains the p value,
which indicates whether the regression coefficient is statistically significantly
different from 0. In Figure 16-4, it is labeled Pr
t|
|
, but it can be called a variety
of other names, including p value, p, and Signif.
In Figure 16-4, the p value for the intercept is shown as 5 49
05
.
e
, which is equal
to 0.0000549 (see the description of scientific notation in Chapter 2). Assuming
we set α at 0.05, the p value is much less than 0.05, so the intercept is statistically
significantly different from zero. But recall that in this example (and usually in
straight-line regression), the intercept doesn’t have any real-world importance.
It’s equals the estimated SBP for a person who weighs 0 kg, which is nonsensical,
so you probably don’t care whether it’s statistically significantly different from
zero or not.