Chapter 4 Data transformation: Assignment
Student name: ____________________________
The following tables gives the GDP (in $US per capita) and life expectancy (in years) for a
group of 20 countries in 2020:
Country Life Expectancy GDP ($US per capita)
Afghanistan 65.17 516.75
Australia 83.20 51680.32
Belgium 80.80 45189.37
Brazil 76.08 6814.88
China 77.10 10408.67
Ecuador 77.22 5627.77
Greece 81.09 17647.23
India 69.89 1933.10
Iran 76.87 2756.75
New Zealand 82.06 41596.51
Papua New Guinea 64.73 2757.01
Philippines 71.36 3301.22
Singapore 83.74 60729.45
Slovak Republic 76.87 19266.51
Somalia 57.70 438.26
South Africa 64.38 5655.87
Spain 82.33 27056.42
Uganda 63.71 822.03
United Kingdom 80.90 41098.08
United States 77.28 63027.68
Response variable: Life Expectancy (variable to be predicted)
1
Explanatory variable: GDP (variable used for prediction)
a We wish to predict life expectancy from GDP. Which would be the response variable,
and which would be the explanatory variable?
b Construct a scatterplot of life expectancy against GDP, and describe the association in
terms of direction, form and strength.
c Use the circle of transformations on page 239 to confirm that the transformations
which might linearise the relationship between life expectancy and GDP are y2, log x
and 1/x.
© Jones, Lipson, Evans, Humberstone, Karakoussis, Staggard 2023 1
Chapter 4 Data transformation: Assignment
2 The y2 transformation:
a Construct a scatterplot of (life expectancy)2 against GDP.
b Fit least squares regression model which enables (life expectancy)2 to be predicted
from GDP. Write down the equation of the model giving the values of the coefficients
rounded to five significant figures.
c Construct a residual plot, and comment on whether the linearity assumption has been
met. show a clear U - shaped pattern
d Determine the value of the coefficient of determination for this model. Give your
answer as a percentage rounded to one decimal place. R^2 = 65.2%
3 The log(x) transformation: y = 20.182 + 7.0757 x log (x)
a Construct a scatterplot of life expectancy against log(GDP), and describe the
association in terms of direction, form and strength.
b Fit least squares regression model which enables life expectancy to be predicted from
log(GDP). Write down the equation of the model giving the values of the coefficients
rounded to five significant figures.
c Construct a residual plot, and comment on whether the linearity assumption has been
met. randomly scattered
d Determine the value of the coefficient of determination for this model. Give your
answer as a percentage rounded to one decimal place.
4 The 1/x transformation: y = 78.372 - 2.4960x 10
a Construct a scatterplot of life expectancy against 1/GDP, and describe the association
in terms of direction, form, and strength.
b Fit least squares regression model which enables life expectancy to be predicted from
1/GDP. Write down the equation of the model giving the values of the coefficients
rounded to five significant figures.
c Construct a residual plot, and comment on whether the linearity assumption
has been met. U - shaped pattern
d Determine the value of the coefficient of determination for this model. Give your
answer as a percentage rounded to one decimal place.
© Jones, Lipson, Evans, Humberstone, Karakoussis, Staggard 2023 2
Chapter 4 Data transformation: Assignment
5
choose log(x) transformation has R^2 is high, a linear residual plot and a strong linear association
a Based on the previous analyses, recommended a regression model for the association
between life expectancy and GDP.
b Use your recommended model to predict, rounded to one decimal place, the life
expectancy of a country where the GDP is:
i. $US20 000 per capita y = 20.182 + 7.0757 x log (20 000) = 50.61
ii. $US80 000 per capita
c Comment on the reliability of your predictions in part b.
© Jones, Lipson, Evans, Humberstone, Karakoussis, Staggard 2023 3
