AP

Statistics
Chapter 9
Re-Expressing data:
Get it Straight
Objective
s:
• Re-expression of data
• Ladder of powers
What is meant by Re-Expressing
Data:
Re-expressing data means making the data
more suitable for analysis by our
methods. ... If it does you can use the 68-
95-99.7 Rule and use the mean and standard
deviation to summarize the data. 3. Another
goal of re-expressing data is to make the
spread of several groups more alike
Why do we need to re-express data

Because we cannot use the terms

"correlation" and "regression" to describe
nonlinear relationships. Therefore, we need
re-express the data in order to describe
the relationships between two
variables.
What are the goals of re-expressing
data?

Goals of Re-expression:
Goal 1:
Make the distribution of a variable more
symmetric.

Goal 2:
Make the spread of several groups more
alike.
What are the goals of re-expressing
data?

Goals of Re-expression:
Goal 3:
Make the form of a scatterplot more nearly
linear.

Goal 4:
Make the scatter in a scatterplot spread out
evenly rather than following a fan shape.
What are the goals of re-expressing
data?

What are the four goals

of re expression?
What are the goals of re-expressing
data?

There are several reasons we may

want to re- express our data:
1) To make the distribution of a
variable more symmetric.
2) To make the spreads of several
groups more alike.
3) To make the form of a
scatterplot more linear.
4) To make the scatter in a
scatterplot more evenly spread
What does a random residual plot
mean?

A residual plot is a graph that

shows the residuals on the
vertical axis and the
independent variable on the
horizontal axis. ...
The first plot shows a random
pattern, indicating a good fit for a
linear model.
What is the ladder of powers?

The Tukey ladder of powers

(sometimes called the Bulging
Rule) is a way to change the
shape of a skewed distribution
so that it becomes normal or
nearly-normal. It can also help to
reduce error variability
(heteroscedasticity). Tukey (1977)
created a table of powers
(numbers to which data can be
raised).
What is the power model?

Power model is a function with

equation of the form y = axn or
a/xn.

Inverse variation power model is a

function with equation y = a/xn (n
> 0). For example,
The relationship between time and
speed, for a fixed distance
traveled, is modeled by an inverse
variation function.
What is the power model?
What is the power model?

Very briefly, a power

model involves taking the
logarithm of both the
dependent and independent
variable. The slope from the
bivariate regression will produce
the power. For an exponential
model, you only take the
logarithm of the dependent
variable.
What type of data are less likely to
benefit from re expression?

Data with positive and

negative values and no
bounds are less likely to benefit
from re-expression.
Straight to the
Point
• We cannot use a linear model unless the
relationship between the two variables is linear.
Often re-expression (transformation) can save
the day, straightening bent relationships so that
we can fit and use a simple linear model.
• Two simple ways to re-express data are with
logarithms and reciprocals.
• Re-expressions can be seen in everyday life—
everybody does it.
Straight to the
Point
• The relationship between fuel efficiency (in
miles per gallon) and weight (in pounds) for late
model cars looks fairly linear at first:
Straight to the
Point
• A look at the residuals plot shows a problem:
 Straight to the
Point
• We can re-express fuel efficiency as gallons per
hundred miles (a reciprocal) and eliminate the
bend in the original scatterplot:
Straight to the
Point
• A look at the residuals plot for the new model
seems more reasonable:
Goals of Re-
expression
• Goal 1: Make the distribution of a variable (as seen in
its histogram, for example) more symmetric.
 It’s easier to summarize the center of a symmetric
distribution, we can use the mean and standard
deviation.
 If the distribution is unimodal also, we can
analysis
using the normal model.
 Here taking the log of the explanatory variable.
Goals of Re-
expression
• Goal 2: Make the spread of several groups (as
seen in side-by-side boxplots) more alike, even
if their centers differ.
 Groups that share a common spread are easier
to compare.
 Here taking the log makes the individual boxplots
more symmetric and gives them spreads that
are more nearly equal.
 Goals of Re-
•expression
Goal 3: Make the form of a scatterplot more
nearly linear.
 Linear scatterplots are easier to model.
 By re-expressing to straighten the scatterplot
relationship we can fit a linear model and use
linear techniques to analysis.
 Here taking the log of the response variable.
 Goals of Re-
•expression
Goal 4: Make the scatter in a scatterplot spread
out evenly rather than thickening at one end.
 Having an even scatter is a condition of many
methods of Statistics, as we will see later.
 This is closely related to goal 2, but often comes
along with goal 3, as seen below. When taking
the log to straighten the data, it also evened out
the spread.
The Ladder of
Powers
• There is a family of simple re-expressions that
move data toward our goals in a consistent
way. This collection of re-expressions is called
the Ladder of Powers.
• The Ladder of Powers orders the effects that
the re-expressions have on data.
 The Ladder of
Powers
Power Name Comment
Square of Try with unimodal distributions that are
2 data values skewed to the left.
Data with positive and negative
1 Raw data values and no bounds are less likely
to benefit from re-expression.
Square root of Counts often benefit from a square
½ data values root re-expression.
We’ll use Measurements that cannot be negative
“0” logarithms here often benefit from a log re-expression.
Reciprocal An uncommon re-expression,
–1/2 square root but sometimes useful.
The reciprocal Ratios of two quantities (e.g., mph)
–1 of the data often benefit from a reciprocal.
The Ladder of
Powers
• The Ladder of Powers orders the effects that
the re-expressions have on data.
• How it works.
 If you try taking the square root of all the values
in a variable and it helps, but not enough, then
move further down the ladder to the log or
reciprocal root. Those re-expressions will have a
similar, but even stronger, effect on your data.
 If you go too far, you can always back up.
 Remember, when you take a negative power,
the direction of the relationship will change. This
is OK, you can always change the sign of the
response variable if you want to keep the same
direction.
Plan B: Attack of the
Logarithms
• When none of the data values is zero or
negative, logarithms can be a helpful ally in the
search for a useful model.
• Try taking the logs of both the x- and y-
variable.
• Then re-express the data using some
combination of x or log(x) vs. y or
log(y).
Plan B: Attack of the
Logarithms
Multiple
Benefits
• We often choose a re-expression for one
reason and then discover that it has helped
other aspects of an analysis.
• For example, a re-expression that makes a
histogram more symmetric might also
straighten a scatterplot or stabilize variance.
Why Not Just Use a
Curve?
• If there’s a curve in the scatterplot, why not just
fit a curve to the data?
Why Not Just Use a
Curve?
• The mathematics and calculations for “curves of
best fit” are considerably more difficult than
“lines of best fit.”
• Besides, straight lines are easy to understand.
 We know how to think about the slope and the y-
intercept.
More Plan B: Modeling
Nonlinear Data - Logarithms
• Two specific types of nonlinear growth.
1. Exponential function (form y = abx)
2. Power function (form y = axb)
• Equations of both forms can be transformed
into linear forms.
• Can then use linear regression to model
and
analyze the transformed data.
• Can also perform an inverse transformation to
obtain a model of the original data.
 To Transform the
exponential Function use
its Inverse the
Logarithmic Function
• Properties of Logarithms


Using Logarithms to Transform
Data

• Logarithms can be useful in straightening a

scatterplot whose data values are greater than
zero.
• Remember, you cannot take the logarithm of a
nonpositive number.
• When you use transformed data to create a
linear model, your regression equation is not in
terms of (x,y) but in terms of the transformed
variable(s) (log ŷ or log x).
Logarithm
Transformations
Example: Testing for
Exponential
• Data Association
View
Scatterplot

• Looks like it has a curved pattern, could

possibly be an exponential
relationship.
Your Turn: Is the following
data exponential & if so,
what is r?
Your Turn: Is the following
data (Hours vs. Number)
exponential & if so, what is
r?
Exponential Regression
Procedure
1. Verify data is exponential.
 Graph scatterplot
 Transform data to linear by taking the log of the
response variable.
2. Calculate the LSRL for the transformed data;
log ŷ =b0+b1x (linear model). Analyze using
linear techniques, LSRL, r, r2, and residuals.
3. Find exponential model for the original data by
inverse transformation of the LSRL,
exponentiating both sides of the LSRL
equation to base 10; ŷ = C • 10kx (exponential
model).
Example:
Data
Annual crude oil production from 1880 to 1970

• Year • Mbbl
1880 30
1890 77
1900 149
1910 328
1920 689
1930 1,412
1940 2,150
1950 3,803
1960 7,674
1970 16,690
What to
do:
1. Graph scatterplot.
2. Transform data to linear (take the log of y).
3. Calculate LSRL of transformed data & graph.
4. Analyze transformed data (r, r2, residual plot).
5. Perform inverse transformation (exponentiate
LSRL to base 10).
6. Graph exponential model.
Back to the
Data
Annual crude oil production from 1880 to 1970

• Year • Mbbl
1880 30
1890 77
1900 149
1910 328
1920 689
1930 1,412
1940 2,150
1950 3,803
1960 7,674
1970 16,690
Models of
Data
• Data is exponential (scatterplot curved pattern
and constant common ratio ≈ 2.1)
• Linear model
 log ŷ=-53.7+.0294x
• Exponential model
 ŷ=(10-53.7) • 10.0294x
 Use model on the calculate to make
predictions,
not the exponential model equation.
 Predict oil production for 1956.
• 6564 Mbbl
 Predict oil production for 1992.
• 75027 Mbbl – extrapolation, be careful.
Your Turn: Exponential
Regression
Models of
Data
• Data is exponential (scatterplot curved pattern
and constant common ratio ≈ 1.5)

• Linear Model
 Log ŷ = -24.11 + .0157x

• Exponential Model
ŷ = (10-24.11) • (10.0157x)
Your Turn: Age vs
Height
Models for
Data
• Data is exponential (scatterplot curved pattern
and constant common ratio ≈ 1.04)
• Linear Model
 Log ŷ = 1.89 + .0241x
• Exponential Model
 ŷ = (101.89) •(10.0214x)
• If comparing Height vs Weight, could a common
ratio be calculated?
 NO, because the explanatory variable Height
does not in crease in equal increments.
 Have to calculate different models and see which
best fits the data.
What does a random residual
plot mean?
A residual plot is a graph that
shows the residuals on the
vertical axis and the
independent variable on the
horizontal axis. ... The first plot
shows a random pattern, indicating
a good fit for a linear model.
 Why is it important to examine a
residual plot even if a scatter plot
appears to be linear?

An examination of the of the residuals often leads

us to discover groups of observations that are
different from the rest.
 How do you create linearity?

Another way to transform the data to achieve linearity

is to take the cube root of the weight values and
graph the cube root of weight versus length.
Note that the resulting scatterplot also has a linear
form.
Can Re-expression remove outliers?

Re-expression can help make data more

symmetric and remove outliers. Can any
relationship be made linear with a re-expression?
Transforming or Re-
Expression
Power Data
Power Function
Model
• Power Function general form: y = axb
• When we apply the log transformation to the response variable
y in an exponential growth model, we produce a linear
relationship. To produce a linear relationship from a power
function model, we apply the log transformation to both
variables (x & y).
• Here is how it is done.
 Power function model: y = axb
 Take the log of both sides of the equation:
log y = log (axb)
 Using the product and power properties of logs, this results
in a linear relationship between log y and log x.
log y = log a +
log xb log y = log a +
b log x
 The power b in the
power function model
 Inverse
•Transformation
Obtaining a power function model for the
original data from the LSRL on the transformed
data.
• LSRL will have the form:
 log ŷ = a + b log x
• Inverse transform the LSRL by
exponentiating both sides of the
equation to base 10.
10log ŷ = 10(a + b log x)
 ŷ = (10a)(10b log x)
 ŷ = (10a)(10log x)b
 ŷ = (10a)(xb) which is in the form y
= C · xb
• A Power Function (can not be done
Power Function
Procedure
1. Graph scatterplot.
2. Determine it is a power function (ie. not
exponential).
3. Transform data to linear (take the log
of y & x).
4. Calculate LSRL of transformed data &
graph.
5. Analyze transformed data (r, r2,
residual plot).
6. Perform inverse transformation (exponentiate
LSRL to base 10).
Example
1
• The table shows the temperature of an instrument
measured as its distance from a heat source is
varied. Find a suitable model for Dist. vs Temp.

• LSRL: log(Temp.) = 4.84 - .255

log(Dist.)
log ŷ = 4.84 - .255 log x
• Power model: Temp. = (104.84)·(Dist.)-.255

ŷ = 104.84 · x-.255
Your
Turn:
• The owner of a Video Game Store records the
business costs and revenue for different
years with the results listed. Find the best
model.

• LSRL: log ŷ = 3.3 + .4 log x

• Power model: ŷ = 103.3 · x.4 or ŷ = (1995)x.4
What Can Go
Wrong?
• Don’t expect your
model to be perfect.

• Don’t stray too far

from the ladder.

• Don’t choose a model

based on R2 alone:
 What Can Go
•
Wrong?
Beware of multiple modes.
 Re-expression cannot pull separate modes
together.
• Watch out for scatterplots that turn around.
 Re-expression can straighten many bent
relationships, but not those that go up then
down, or down then up.
What Can Go
Wrong?
• Watch out for negative data values.
 It’s impossible to re-express negative values by
any power that is not a whole number on the
Ladder of Powers or to re-express values that
are zero for negative powers.
• Watch for data far from 1.
 Data values that are all very far from 1 may not
be much affected by re-expression unless the
range is very large. If all the data values are
large (e.g., years), consider subtracting a
constant to bring them back near 1.
What have we
learned?
• When the conditions for regression are not met,
a simple re-expression of the data may help.
• A re-expression may make the:
 Distribution of a variable more symmetric.
 Spread across different groups more similar.
 Form of a scatterplot straighter.
 Scatter around the line in a scatterplot more
consistent.
What have we
learned?
• Taking logs is often a good, simple starting
point.
 To search further, the Ladder of Powers or the
log-log approach can help us find a good re-
expression.
• Our models won’t be perfect, but re-expression
can lead us to a useful model.