Stat 301 – Lecture 8
Regression model
Y   y| x  
•Y
represents a value of the response
variable.
• y| x represents the population mean
response for a given value of the
explanatory variable, x.
• represents the random error
1
Linear Regression Model
Y   y| x     0  1 x  
0
The Y-intercept parameter.
1
The slope parameter.
2
Residual
(Observed Y – Fitted Y)
Fit
Yˆ  ˆ y| x  ˆ0  ˆ1 x
Residual

ˆ  Y  Yˆ  Y  ˆ0  ˆ1 x

3
1
Stat 301 – Lecture 8
Conditions
The relationship is linear.
The random error term,  , is
 Independent
 Identically
distributed
 Normally distributed with
standard deviation,  .
4
Residual vs. Explanatory
0.3
0.2
Residual
0.1
0.0
-0.1
-0.2
-0.3
300
350
400
CO2
5
Residual vs. Predicted
6
2
Stat 301 – Lecture 8
Interpretation
Random scatter around the
zero line indicates that the
linear model is adequate for
the relationship between
carbon dioxide and
temperature.
7
Patterns
Over/Under/Over or
Under/Over/Under
The linear model may not be
adequate. We could do
better by accounting for
curvature with a different
model.
8
Speed and Stopping Distance
9
3
Stat 301 – Lecture 8
Patterns
Two, or more, groups
May
require separate
regression models for each
group.
10
Gas used vs. Temperature
11
Checking Conditions
Independence.
Hard
to check this but the
fact that we obtained the
data through a random
sample of years assures us
that the statistical methods
should work.
12
4
Stat 301 – Lecture 8
Checking Conditions
Identically distributed.
 Check
using an outlier box plot.
Unusual points may come from
a different distribution
 Check using a histogram. Bimodal shape could indicate two
different distributions.
13
Checking Conditions
Normally distributed.
Check
with a histogram.
Symmetric and mounded in
the middle.
Check with a normal
quantile plot. Points falling
close to a diagonal line.
14
Distributions
3
.99
2
.95
.90
1
.75
Normal Quantile Plot
Residuals Temp
0
.50
.25
-1
.10
.05
-2
.01
-3
10
6
Count
8
4
2
-0.2
-0.1
0
0.1
0.2
0.3
15
5
Stat 301 – Lecture 8
Residuals
Histogram is skewed right and
mounded to the left of zero.
Box plot is skewed right with no
unusual points.
Normal quantile plot has points
that do not follow the diagonal,
normal model, line very well.
16
Checking Conditions
Constant variance.
Check
the plot of residuals
versus the explanatory or
predicted.
Points should show the
same spread for all values of
the explanatory variable.
17
Residual
Non-constant variance
Explanatory or Predicted
18
6
Stat 301 – Lecture 8
Residual vs. Explanatory
0.3
0.2
Residual
0.1
0.0
-0.1
-0.2
-0.3
300
350
400
CO2
19
Residual vs. Predicted
20
Constant Variance
Points show about the same
amount of spread for all
values of the explanatory
variable.
21
7
Stat 301 – Lecture 8
Conclusion
The independence,
identically distributed and
common variance conditions
appear to be satisfied.
The normal distribution
condition may not be met for
these data.
22
Consequences
The P-values for tests may
not be correct.
However, the P-value was
so small, there is still strong
evidence for a linear
relationship between carbon
dioxide and temperature.
23
Consequences
The stated confidence level
may not give the true
coverage rate.
We have confidence in the
intervals, it may not be 95%.
24
8