Graphics Reference
In-Depth Information
. We have options in the
ancova
function to show any one of
a) Horizontal slopes (ANOVA ignoring the covariate),
b) Common slope (ordinary ANCOVA),
c) Common intercept (regression ignoring the factor),
d) Distinct lines (interaction of factor and covariate).
Example - Hot Dog Data
6.5.1
Hot dogs based on poultry are said to be healthier than ones made from either meat
(beefandpork)orallbeef.Abasisforthisclaimmaybethelower-calorie(fat)content
of poultry hot dogs. Is this advantage of poultry hot dogs offset by a higher sodium
content than meat hot dogs?
Researchers for Consumer Reports analyzed three types of hot dog: beef, poultry,
and meat (mostly pork and beef, but up to
% poultry meat). he data available
in file (hh/datasets/hotdog.dat) come from Consumers Union (
) and were later
used by Moore and McCabe (
):
Type:
type of hot dog (beef, meat, or poultry)
Calories:
calories per hot dog
Sodium:
milligrams of sodium per hot dog
We used these data to model
Sodium
as a function of the categorical variable
Type
and the continuous variable
Calories
.
he model displayed in Fig.
.
has common slope and possibly differing inter-
cepts. Displays comparable to Fig.
.
can be constructed for the other three mod-
els. he common slope model is the correct final model, a conclusion that required
support from the ANOVA table for the interaction model. In context, for given fat
(calorie) content,
poultry
hot dogs contain significantly more sodium than
beef
or
meat
hot dogs, but the three
type
s have insignificantly different increases in
sodium as calories increase.
Figure
.
.
he let set of panels show one level each. he right panel, labeled “superpose,” shows all
three groups. Sometimes it is easier to interpret the individual panels, other times the superposed
panel. We therefore normally print both. he HH Heiberger and Holland (
)
ancova
function
automatically constructs all panels from a single call (Files: hh/regbb/code/hotdog.s,
hh/csc/code/hotdog.csc.s)
