3 Suppose the relationship between applied stress x and time-to-failure y is
Example 12.3 Suppose the relationship between applied stress x and time-to-failure y is
described by the simple linear regression model with true regression line y 5 65 2 1.2x and s58 . Then for any fixed value x of stress, time-to-failure has a normal distribution with mean value
65 2 1.2x and standard deviation 8.
Roughly speaking, in the population consisting of all (x, y) points, the magnitude of a typical deviation from the true regression line is about 8. For x 5 20 , Y has
mean value m Y 20 5 65 2 1.2(20) 5 41, so
P(Y . 50 when x 5 20) 5 P aZ .
8 b 5 1 2 (1.13) 5 .1292
The probability that time-to-failure exceeds 50 when applied stress is 25 is, because
m Y 25 5 35,
P(Y . 50 when x 5 25) 5 P aZ .
8 b 5 1 2 (1.88) 5 .0301
These probabilities are illustrated as the shaded areas in Figure 12.5.
y
P(Y
50 when x 20) .1292
P(Y
50 when x 25) .0301
35 True regression line
Figure 12.5 Probabilities based on the simple linear regression model
Suppose that Y 1 denotes an observation on time-to-failure made with x 5 25
and Y 2 denotes an independent observation made with x 5 24 . Then Y 1 2Y 2 is nor- mally distributed with mean value E(Y 1 2Y 2 )5b 1 5 21.2 , variance V(Y 1 2Y 2 )5s 2 1s 2 5 128 , and standard deviation 1128 5 11.314 . The prob- ability that Y 1 exceeds Y 2 is
12.1 The Simple Linear Regression Model
b 5 P(Z . .11) 5 .4562
That is, even though we expected Y to decrease when x increases by 1 unit, it is not unlikely that the observed Y at x11 will be larger than the observed Y at x.
■
EXERCISES Section 12.1 (1–11)
1. The efficiency ratio for a steel specimen immersed in a phos-
Construct scatter plots of NO x emissions versus age. What
phating tank is the weight of the phosphate coating divided
appears to be the nature of the relationship between these
by the metal loss (both in mgft 2 ). The article “Statistical
two variables? [Note: The authors of the cited article com-
Process Control of a Phosphate Coating Line” (Wire J. Intl.,
mented on the relationship.]
May 1997: 78–81) gave the accompanying data on tank tem-
3. Bivariate data often arises from the use of two different tech-
perature (x) and efficiency ratio (y).
niques to measure the same quantity. As an example, the
accompanying observations on x5 hydrogen concentration
Ratio
.84 1.31 1.42 1.03 1.07 1.08 1.04 (ppm) using a gas chromatography method and y5 concen- tration using a new sensor method were read from a graph in
the article “A New Method to Measure the Diffusible
Ratio
Hydrogen Content in Steel Weldments Using a Polymer Electrolyte-Based Hydrogen Sensor” (Welding Res., July
1997: 251s–256s).
a. Construct stem-and-leaf displays of both temperature and
x
efficiency ratio, and comment on interesting features.
b. Is the value of efficiency ratio completely and uniquely
y
determined by tank temperature? Explain your reasoning.
c. Construct a scatter plot of the data. Does it appear that
Construct a scatter plot. Does there appear to be a very strong
efficiency ratio could be very well predicted by the value
relationship between the two types of concentration meas-
of temperature? Explain your reasoning.
urements? Do the two methods appear to be measuring
2. The article “Exhaust Emissions from Four-Stroke Lawn
roughly the same quantity? Explain your reasoning.
Mower Engines” (J. of the Air and Water Mgmnt. Assoc.,
4. A study to assess the capability of subsurface flow wetland sys-
1997: 945–952) reported data from a study in which both a
tems to remove biochemical oxygen demand (BOD) and vari-
baseline gasoline mixture and a reformulated gasoline were
ous other chemical constituents resulted in the accompanying
used. Consider the following observations on age (yr) and
data on x 5 BOD mass loading (kghad) and y 5 BOD mass
NO x emissions (gkWh):
removal (kghad) (“Subsurface Flow Wetlands—A Performance Evaluation,” Water Envir. Res., 1995: 244–247).
6 7 8 9 10 a. Construct boxplots of both mass loading and mass
Age
16 9 0 12 4 removal, and comment on any interesting features.
1.24 b. Construct a scatter plot of the data, and comment on any
1.42 interesting features.
CHAPTER 12 Simple Linear Regression and Correlation
5. The article “Objective Measurement of the Stretchability of
ered regressing y 5 28-day standard-cured strength (psi)
Mozzarella Cheese” (J. of Texture Studies, 1992: 185–194)
against x 5 accelerated strength (psi) . Suppose the equation
reported on an experiment to investigate how the behavior of
of the true regression line is y 5 1800 1 1.3x .
mozzarella cheese varied with temperature. Consider the
a. What is the expected value of 28-day strength when accel-
accompanying data on x 5 temperature and y 5 elongation
erated strength
() at failure of the cheese. [Note: The researchers were
b. By how much can we expect 28-day strength to change
Italian and used real mozzarella cheese, not the poor cousin
when accelerated strength increases by 1 psi?
widely available in the United States.]
c. Answer part (b) for an increase of 100 psi. d. Answer part (b) for a decrease of 100 psi.
x
59 63 68 72 74 78 83 8. Referring to Exercise 7, suppose that the standard deviation
of the random deviation is 350 psi. P
a. What is the probability that the observed value of 28-day strength will exceed 5000 psi when the value of acceler- ated strength is 2000?
a. Construct a scatter plot in which the axes intersect at
b. Repeat part (a) with 2500 in place of 2000.
(0, 0). Mark 0, 20, 40, 60, 80, and 100 on the horizontal
c. Consider making two independent observations on 28-day
axis and 0, 50, 100, 150, 200, and 250 on the vertical
strength, the first for an accelerated strength of 2000 and
axis.
the second for x 5 2500 . What is the probability that the
b. Construct a scatter plot in which the axes intersect at
second observation will exceed the first by more than
(55, 100), as was done in the cited article. Does this
1000 psi?
plot seem preferable to the one in part (a)? Explain
d. Let Y 1 and Y 2 denote observations on 28-day strength when
your reasoning.
x5x and x5x , respectively. By how much would x
c. What do the plots of parts (a) and (b) suggest about the
have to exceed x 1 in order that P(Y 2 .Y 1 ) 5 .95 ?
nature of the relationship between the two variables?
9. The flow rate y (m 3 min) in a device used for air-quality
6. One factor in the development of tennis elbow, a malady that
measurement depends on the pressure drop x (in. of water)
strikes fear in the hearts of all serious tennis players, is the
across the device’s filter. Suppose that for x values between
impact-induced vibration of the racket-and-arm system at ball
5 and 20, the two variables are related according to the simple
contact. It is well known that the likelihood of getting tennis
linear regression model with true regression line
elbow depends on various properties of the racket used.
y 5 2.12 1 .095x .
Consider the scatter plot of x5 racket resonance frequency
a. What is the expected change in flow rate associated with
(Hz) and y 5 sum of peak-to-peak acceleration (a character-
a 1-in. increase in pressure drop? Explain.
istic of arm vibration, in msecsec) for n 5 23 different rack-
b. What change in flow rate can be expected when pressure
ets (“Transfer of Tennis Racket Vibrations into the Human
drop decreases by 5 in.?
Forearm,” Medicine and Science in Sports and Exercise,
c. What is the expected flow rate for a pressure drop of
1992: 1134–1140). Discuss interesting features of the data
10 in.? A drop of 15 in.?
and scatter plot.
d. Suppose s 5 .025 and consider a pressure drop of 10 in. What is the probability that the observed value of flow rate
y
will exceed .835? That observed flow rate will exceed .840? e. What is the probability that an observation on flow rate
when pressure drop is 10 in. will exceed an observation on
flow rate made when pressure drop is 11 in.?
10. Suppose the expected cost of a production run is related to the
size of the run by the equation
. Let Y denote an observation on the cost of a run. If the variables’ size and cost
y 5 4000 1 10x
are related according to the simple linear regression model,
could it be the case that P(Y . 50 when x 5 100) 5 .05
and ? Explain.
P(Y . 6500 when x 5 200) 5 .10
11. Suppose that in a certain chemical process the reaction time y (hr) is related to the temperature (°F ) in the chamber in
22 x
which the reaction takes place according to the simple lin-
ear regression model with equation y 5 5.00 2 .01x and s 5 .075 .
7. The article “Some Field Experience in the Use of an
a. What is the expected change in reaction time for a 1°F
Accelerated Method in Estimating 28-Day Strength of
increase in temperature? For a 10°F increase in
Concrete” (J. of Amer. Concrete Institute, 1969: 895) consid-
temperature?
12.2 Estimating Model Parameters
b. What is the expected reaction time when temperature is
d. What is the probability that two independently observed
200°F? When temperature is 250°F?
reaction times for temperatures 1° apart are such that the
c. Suppose five observations are made independently on reac-
time at the higher temperature exceeds the time at the
tion time, each one for a temperature of 250°F. What is the
lower temperature?
probability that all five times are between 2.4 and 2.6 hr?