R Help 9: Data Transformations

R Help 9: Data Transformations

R Help

Word recall (log-transforming a predictor)

  • Load the wordrecall data.
  • Fit a simple linear regression model of prop on time.
  • Display a scatterplot of the data and add the regression line.
  • Display a residual plot with fitted (predicted) values on the horizontal axis.
  • Display a normal probability plot of the residuals and add a diagonal line to the plot. The argument "datax" determines which way round to plot the axes (false by default, which plots the data on the vertical axis, or true, which plots the data on the horizontal axis).
  • Load the nortest package to access the Anderson-Darling normality test.
  • Create a log(time) variable and fit a simple linear regression model of prop on log(time).
  • Repeat diagnostic plots and normality tests.
  • Create a prop^-1.25 variable and fit a simple linear regression model of prop^-1.25 on time.
  • Repeat diagnostic plots and normality tests.
  • Use prop on log(time) model to find:
    • 95% prediction interval for a prop at time 1000.
    • 95% confidence interval for the expected change in prop for a 10-fold increase in time.
wordrecall <- read.table("~/path-to-folder/wordrecall.txt", header=T)
attach(wordrecall)

model.1 <- lm(prop ~ time)
summary(model.1)
#               Estimate Std. Error t value Pr(>|t|)    
# (Intercept)  5.259e-01  4.881e-02  10.774 3.49e-07 ***
# time        -5.571e-05  1.457e-05  -3.825  0.00282 ** 
# Multiple R-squared:  0.5709,  Adjusted R-squared:  0.5318 

plot(x=time, y=prop, ylim=c(-0.1, 0.9),
     panel.last = lines(sort(time), fitted(model.1)[order(time)]))

plot(x=fitted(model.1), y=residuals(model.1),
     panel.last = abline(h=0, lty=2))

qqnorm(residuals(model.1), main="", datax=TRUE)
qqline(residuals(model.1), datax=TRUE)

library(nortest)
ad.test(residuals(model.1)) # A = 0.262, p-value = 0.6426

lntime <- log(time)

model.2 <- lm(prop ~ lntime)
summary(model.2)
#              Estimate Std. Error t value Pr(>|t|)    
# (Intercept)  0.846415   0.014195   59.63 3.65e-15 ***
# lntime      -0.079227   0.002416  -32.80 2.53e-12 ***
# ---
# Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
# 
# Residual standard error: 0.02339 on 11 degrees of freedom
# Multiple R-squared:  0.9899,  Adjusted R-squared:  0.989 
# F-statistic:  1076 on 1 and 11 DF,  p-value: 2.525e-12

plot(x=lntime, y=prop,
     panel.last = lines(sort(lntime), fitted(model.2)[order(lntime)]))

plot(x=fitted(model.2), y=residuals(model.2),
     panel.last = abline(h=0, lty=2))

qqnorm(residuals(model.2), main="", datax=TRUE)
qqline(residuals(model.2), datax=TRUE)

ad.test(residuals(model.2)) # A = 0.3216, p-value = 0.4869

prop1.25 <- prop^-1.25

model.3 <- lm(prop1.25 ~ time)
summary(model.3)
#              Estimate Std. Error t value Pr(>|t|)    
# (Intercept) 1.8693698  0.3869678   4.831 0.000527 ***
# time        0.0019708  0.0001155  17.067 2.91e-09 ***
# Multiple R-squared:  0.9636,  Adjusted R-squared:  0.9603 

plot(x=time, y=prop1.25,
     panel.last = lines(sort(time), fitted(model.3)[order(time)]))

plot(x=fitted(model.3), y=residuals(model.3),
     panel.last = abline(h=0, lty=2))

qqnorm(residuals(model.3), main="", datax=TRUE)
qqline(residuals(model.3), datax=TRUE)

ad.test(residuals(model.3)) # A = 1.191, p-value = 0.002584

predict(model.2, interval="prediction",
        newdata=data.frame(lntime=log(1000)))
#         fit       lwr       upr
# 1 0.2991353 0.2449729 0.3532978

confint(model.2)[2,]*log(10) # 95% CI for 10-fold increase in time
#      2.5 %     97.5 % 
# -0.1946689 -0.1701845 

detach(wordrecall)

Mammal gestation (log-transforming the response)

  • Load the mammgest data.
  • Fit a simple linear regression model of Gestation on Birthwgt.
  • Display a scatterplot of the data and add the regression line.
  • Display residual plots with fitted (predicted) values on the horizontal axis.
  • Display a normal probability plot of the residuals and add a diagonal line to the plot.
  • Apply the Anderson-Darling normality test using the nortest package.
  • Create a log(Gestation) variable and fit a simple linear regression model of log(Gestation) on Birthwgt.
  • Repeat diagnostic plots and normality tests.
  • Use log(Gestation) on Birthwgt model to find:
    • 95% prediction interval for Gestation for a Birthwgt of 50.
    • 95% confidence interval for proportional change in median Gestation for a 1-pound increase in Birthwgt.
    • 95% confidence interval for proportional change in median Gestation for a 10-pound increase in Birthwgt.
mammgest <- read.table("~/path-to-folder/mammgest.txt", header=T)
attach(mammgest)

model.1 <- lm(Gestation ~ Birthwgt)
summary(model.1)
#             Estimate Std. Error t value Pr(>|t|)    
# (Intercept) 187.0837    26.9426   6.944 6.73e-05 ***
# Birthwgt      3.5914     0.5247   6.844 7.52e-05 ***
# Multiple R-squared:  0.8388,  Adjusted R-squared:  0.8209 

plot(x=Birthwgt, y=Gestation,
     panel.last = lines(sort(Birthwgt), fitted(model.1)[order(Birthwgt)]))

plot(x=fitted(model.1), y=residuals(model.1),
     panel.last = abline(h=0, lty=2))

qqnorm(residuals(model.1), main="", datax=TRUE)
qqline(residuals(model.1), datax=TRUE)

ad.test(residuals(model.1)) # A = 0.3116, p-value = 0.503

lnGest <- log(Gestation)

model.2 <- lm(lnGest ~ Birthwgt)
summary(model.2)
#             Estimate Std. Error t value Pr(>|t|)    
# (Intercept) 5.278817   0.088177  59.866  5.1e-13 ***
# Birthwgt    0.010410   0.001717   6.062 0.000188 ***
# ---
# Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
# 
# Residual standard error: 0.2163 on 9 degrees of freedom
# Multiple R-squared:  0.8033,  Adjusted R-squared:  0.7814 
# F-statistic: 36.75 on 1 and 9 DF,  p-value: 0.0001878

plot(x=Birthwgt, y=lnGest,
     panel.last = lines(sort(Birthwgt), fitted(model.2)[order(Birthwgt)]))

plot(x=fitted(model.2), y=residuals(model.2),
     panel.last = abline(h=0, lty=2))

qqnorm(residuals(model.2), main="", datax=TRUE)
qqline(residuals(model.2), datax=TRUE)

ad.test(residuals(model.2)) # A = 0.3135, p-value = 0.4963

exp(predict(model.2, interval="prediction",
            newdata=data.frame(Birthwgt=50)))
#        fit      lwr      upr
# 1 330.0781 197.3013 552.2092

# proportional change in median gestation for 1-unit increase in birthwgt
exp(coefficients(model.2)[2]) # 1.010465
exp(confint(model.2)[2,]) # 95% CI
#    2.5 %   97.5 % 
# 1.006547 1.014398 

# proportional change in median gestation for 10-unit increase in birthwgt
exp(10*coefficients(model.2)[2]) # 1.109714
exp(10*confint(model.2)[2,]) # 95% CI
#    2.5 %   97.5 % 
# 1.067429 1.153674 

detach(mammgest)

Shortleaf pine trees (log-transforming both response and predictor)

  • Load the shortleaf data.
  • Fit a simple linear regression model of Vol on Diam.
  • Display a scatterplot of the data and add the regression line.
  • Display residual plots with fitted (predicted) values on the horizontal axis.
  • Display a normal probability plot of the residuals and add a diagonal line to the plot.
  • Apply the Anderson-Darling normality test using the nortest package.
  • Create a log(Diam) variable and fit a simple linear regression model of Vol on log(Diam).
  • Repeat diagnostic plots and normality tests.
  • Create a log(Vol) variable and fit a simple linear regression model of log(Vol) on log(Diam).
  • Repeat diagnostic plots and normality tests.
  • Use log(Vol) on log(Diam) model to find:
    • 95% confidence interval for median Vol for a Diam of 10.
    • 95% confidence interval for proportional change in median Vol for a 2-fold increase in Diam.
shortleaf <- read.table("~/path-to-folder/shortleaf.txt", header=T)
attach(shortleaf)

model.1 <- lm(Vol ~ Diam)
summary(model.1)
#             Estimate Std. Error t value Pr(>|t|)    
# (Intercept) -41.5681     3.4269  -12.13   <2e-16 ***
# Diam          6.8367     0.2877   23.77   <2e-16 ***
# ---
# Multiple R-squared:  0.8926,  Adjusted R-squared:  0.891 

plot(x=Diam, y=Vol,
     panel.last = lines(sort(Diam), fitted(model.1)[order(Diam)]))

plot(x=fitted(model.1), y=residuals(model.1),
     panel.last = abline(h=0, lty=2))

qqnorm(residuals(model.1), main="", datax=TRUE)
qqline(residuals(model.1), datax=TRUE)

ad.test(residuals(model.1)) # A = 0.9913, p-value = 0.01215

lnDiam <- log(Diam)

model.2 <- lm(Vol ~ lnDiam)
summary(model.2)
#             Estimate Std. Error t value Pr(>|t|)    
# (Intercept) -116.162     10.830  -10.73 2.88e-16 ***
# lnDiam        64.536      4.562   14.15  < 2e-16 ***
# Multiple R-squared:  0.7464,  Adjusted R-squared:  0.7427 

plot(x=lnDiam, y=Vol,
     panel.last = lines(sort(lnDiam), fitted(model.2)[order(lnDiam)]))

plot(x=fitted(model.2), y=residuals(model.2),
     panel.last = abline(h=0, lty=2))

qqnorm(residuals(model.2), main="", datax=TRUE)
qqline(residuals(model.2), datax=TRUE)

ad.test(residuals(model.2)) # A = 2.3845, p-value = 4.273e-06

lnVol <- log(Vol)

model.3 <- lm(lnVol ~ lnDiam)
summary(model.3)
#             Estimate Std. Error t value Pr(>|t|)    
# (Intercept)  -2.8718     0.1216  -23.63   <2e-16 ***
# lnDiam        2.5644     0.0512   50.09   <2e-16 ***
# ---
# Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
# 
# Residual standard error: 0.1703 on 68 degrees of freedom
# Multiple R-squared:  0.9736,  Adjusted R-squared:  0.9732 
# F-statistic:  2509 on 1 and 68 DF,  p-value: < 2.2e-16

plot(x=lnDiam, y=lnVol,
     panel.last = lines(sort(lnDiam), fitted(model.3)[order(lnDiam)]))

plot(x=fitted(model.3), y=residuals(model.3),
     panel.last = abline(h=0, lty=2))

qqnorm(residuals(model.3), main="", datax=TRUE)
qqline(residuals(model.3), datax=TRUE)

ad.test(residuals(model.3)) # A = 0.5309, p-value = 0.1692

exp(predict(model.3, interval="confidence",
            newdata=data.frame(lnDiam=log(10))))
#        fit      lwr      upr
# 1 20.75934 19.92952 21.62372

# proportional change in median Vol for 2-fold increase in Diam
2^(coefficients(model.3)[2]) # 5.915155
2^(confint(model.3)[2,]) # 95% CI
#    2.5 %   97.5 % 
# 5.510776 6.349207

detach(shortleaf)

Underground air quality (interactions)

  • Load the swallows data.
  • Load the car package to access 3D scatterplots.
  • Create interaction variables and fit a multiple linear regression model of Vent on O2 + CO2 + Type + TypeO2 + TypeCO2 + CO2O2.
  • Use anova function to display the ANOVA table with sequential (type I) sums of squares.
  • Calculate partial F-statistic and p-value.
  • Display residual plots with fitted (predicted) values on the horizontal axis.
  • Fit a multiple linear regression model of Vent on O2 + CO2 + Type.
  • Display residual plots with fitted (predicted) values on the horizontal axis.
  • Display a normal probability plot of the residuals and add a diagonal line to the plot.
  • Apply the Anderson-Darling normality test using the nortest package.
swallows <- read.table("~/path-to-folder/allswallows.txt", header=T)
attach(swallows)

library(car)
scatter3d(Vent ~ O2 + CO2, subset=Type==1) # adult
scatter3d(Vent ~ O2 + CO2, subset=Type==0) # nestling
scatter3d(Vent ~ O2 + CO2, subset=Type==0, revolutions=3, speed=0.5, grid=F)

TypeO2 <- Type*O2
TypeCO2 <- Type*CO2
CO2O2 <- CO2*O2

model.1 <- lm(Vent ~ O2 + CO2 + Type + TypeO2 + TypeCO2 + CO2O2)
summary(model.1)
#             Estimate Std. Error t value Pr(>|t|)  
# (Intercept)  -18.399    160.007  -0.115   0.9086  
# O2             1.189      9.854   0.121   0.9041  
# CO2           54.281     25.987   2.089   0.0378 *
# Type         111.658    157.742   0.708   0.4797  
# TypeO2        -7.008      9.560  -0.733   0.4642  
# TypeCO2        2.311      7.126   0.324   0.7460  
# CO2O2         -1.449      1.593  -0.909   0.3642  
# ---
# Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
# 
# Residual standard error: 165.6 on 233 degrees of freedom
# Multiple R-squared:  0.272,  Adjusted R-squared:  0.2533 
# F-statistic: 14.51 on 6 and 233 DF,  p-value: 4.642e-14

anova(model.1) # Sequential (type I) SS
#            Df  Sum Sq Mean Sq F value  Pr(>F)    
# O2          1   93651   93651  3.4156 0.06585 .  
# CO2         1 2247696 2247696 81.9762 < 2e-16 ***
# Type        1    5910    5910  0.2156 0.64288    
# TypeO2      1   14735   14735  0.5374 0.46425    
# TypeCO2     1    2884    2884  0.1052 0.74598    
# CO2O2       1   22664   22664  0.8266 0.36421    
# Residuals 233 6388603   27419                    
# ---
# Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

((14735+2884+22664)/3)/27419 # F-stat = 0.4897212
pf(0.49, 3, 233, lower.tail=F) # p-value = 0.6895548

plot(x=fitted(model.1), y=residuals(model.1),
     panel.last = abline(h=0, lty=2))

model.2 <- lm(Vent ~ O2 + CO2 + Type)
summary(model.2)
#             Estimate Std. Error t value Pr(>|t|)    
# (Intercept)  136.767     79.334   1.724    0.086 .  
# O2            -8.834      4.765  -1.854    0.065 .  
# CO2           32.258      3.551   9.084   <2e-16 ***
# Type           9.925     21.308   0.466    0.642    
# ---
# Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
# 
# Residual standard error: 165 on 236 degrees of freedom
# Multiple R-squared:  0.2675,  Adjusted R-squared:  0.2581 
# F-statistic: 28.72 on 3 and 236 DF,  p-value: 7.219e-16

plot(x=fitted(model.2), y=residuals(model.2),
     panel.last = abline(h=0, lty=2))

qqnorm(residuals(model.2), main="", datax=TRUE)
qqline(residuals(model.2), datax=TRUE)

ad.test(residuals(model.2)) # A = 0.3175, p-value = 0.5358

detach(swallows)

Bluegill fish (polynomial regression)

  • Load the bluegills data.
  • Display a scatterplot of the data.
  • Create an age-squared variable and fit a multiple linear regression model of length on age + agesq.
  • Add a quadratic regression line to the scatterplot.
  • Find a 95% prediction interval for length for an age of 5.
bluegills <- read.table("~/path-to-folder/bluegills.txt", header=T)
attach(bluegills)

plot(x=age, y=length)

agesq <- age^2

model <- lm(length ~ age + agesq)
summary(model)
#             Estimate Std. Error t value Pr(>|t|)    
# (Intercept)   13.622     11.016   1.237     0.22    
# age           54.049      6.489   8.330 2.81e-12 ***
# agesq         -4.719      0.944  -4.999 3.67e-06 ***
# ---
# Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
# 
# Residual standard error: 10.91 on 75 degrees of freedom
# Multiple R-squared:  0.8011,  Adjusted R-squared:  0.7958 
# F-statistic: 151.1 on 2 and 75 DF,  p-value: < 2.2e-16

newX <- seq(min(age), max(age), length=100)
newXsq <- newX**2

plot(x=age, y=length,
     panel.last = lines(newX,
                        predict(model,
                                newdata=data.frame(age=newX, agesq=newXsq))))

predict(model, interval="prediction",
        newdata=data.frame(age=5, agesq=25))
#        fit     lwr      upr
# 1 165.9023 143.487 188.3177

detach(bluegills)

Experiment yield (polynomial regression)

  • Load the yield data.
  • Fit a simple linear regression model of Yield on Temp.
  • Display a scatterplot of the data and add the simple linear regression line.
  • Create a Temp-squared variable and fit a multiple linear regression model of Yield on Temp + Tempsq.
  • Display a scatterplot of the data and add a quadratic regression line to the scatterplot.
yield <- read.table("~/path-to-folder/yield.txt", header=T)
attach(yield)

model.1 <- lm(Yield ~ Temp)
summary(model.1)
#             Estimate Std. Error t value Pr(>|t|)    
# (Intercept) 2.306306   0.469075   4.917 0.000282 ***
# Temp        0.006757   0.005873   1.151 0.270641    
# ---
# Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
# 
# Residual standard error: 0.3913 on 13 degrees of freedom
# Multiple R-squared:  0.09242,  Adjusted R-squared:  0.0226 
# F-statistic: 1.324 on 1 and 13 DF,  p-value: 0.2706

plot(x=Temp, y=Yield,
     panel.last = lines(sort(Temp), fitted(model.1)[order(Temp)]))

Tempsq <- Temp^2

model.2 <- lm(Yield ~ Temp + Tempsq)
summary(model.2)
#               Estimate Std. Error t value Pr(>|t|)    
# (Intercept)  7.9604811  1.2589183   6.323 3.81e-05 ***
# Temp        -0.1537113  0.0349408  -4.399 0.000867 ***
# Tempsq       0.0010756  0.0002329   4.618 0.000592 ***
# ---
# Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
# 
# Residual standard error: 0.2444 on 12 degrees of freedom
# Multiple R-squared:  0.6732,  Adjusted R-squared:  0.6187 
# F-statistic: 12.36 on 2 and 12 DF,  p-value: 0.001218

newX <- seq(min(Temp), max(Temp), length=100)
newXsq <- newX**2

plot(x=Temp, y=Yield,
     panel.last = lines(newX,
                        predict(model.2,
                                newdata=data.frame(Temp=newX, Tempsq=newXsq))))

detach(yield)

Chemical odor (polynomial regression)

  • Load the odor data.
  • Create squared variables and fit a multiple linear regression model of Odor on Temp + Ratio + Height + Tempsq + Ratiosq + Heightsq.
  • Fit a multiple linear regression model of Odor on Temp + Ratio + Height + Tempsq + Ratiosq.
odor <- read.table("~/path-to-folder/odor.txt", header=T)
attach(odor)

Tempsq <- Temp^2
Ratiosq <- Ratio^2
Heightsq <- Height^2

model.1 <- lm(Odor ~ Temp + Ratio + Height + Tempsq + Ratiosq + Heightsq)
summary(model.1)
#             Estimate Std. Error t value Pr(>|t|)   
# (Intercept)  -30.667     10.840  -2.829   0.0222 * 
# Temp         -12.125      6.638  -1.827   0.1052   
# Ratio        -17.000      6.638  -2.561   0.0336 * 
# Height       -21.375      6.638  -3.220   0.0122 * 
# Tempsq        32.083      9.771   3.284   0.0111 * 
# Ratiosq       47.833      9.771   4.896   0.0012 **
# Heightsq       6.083      9.771   0.623   0.5509   
# ---
# Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
# 
# Residual standard error: 18.77 on 8 degrees of freedom
# Multiple R-squared:  0.8683,  Adjusted R-squared:  0.7695 
# F-statistic: 8.789 on 6 and 8 DF,  p-value: 0.003616

model.2 <- lm(Odor ~ Temp + Ratio + Height + Tempsq + Ratiosq)
summary(model.2)
#             Estimate Std. Error t value Pr(>|t|)    
# (Intercept)  -26.923      8.707  -3.092 0.012884 *  
# Temp         -12.125      6.408  -1.892 0.091024 .  
# Ratio        -17.000      6.408  -2.653 0.026350 *  
# Height       -21.375      6.408  -3.336 0.008720 ** 
# Tempsq        31.615      9.404   3.362 0.008366 ** 
# Ratiosq       47.365      9.404   5.036 0.000703 ***
# ---
# Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
# 
# Residual standard error: 18.12 on 9 degrees of freedom
# Multiple R-squared:  0.8619,  Adjusted R-squared:  0.7852 
# F-statistic: 11.23 on 5 and 9 DF,  p-value: 0.001169

detach(odor)

Legend
[1]Link
Has Tooltip/Popover
 Toggleable Visibility