Regression 1 practice

Is there evidence of a relationship between height and mass for a

random sample of APES 2024 honours students?
90

80

70
mass

60

50

40
1.5 1.6 1.7 1.8 1.9
height

1
## # A tibble: 10 x 4
## ‘height[1:10]‘ ‘mass[1:10]‘ ‘height[11:20]‘ ‘mass[11:20]‘
## <dbl> <dbl> <dbl> <dbl>
## 1 1.72 67 1.75 64
## 2 1.66 66 1.55 65
## 3 1.59 67 1.61 75
## 4 1.5 42 1.65 70
## 5 1.78 83 1.61 55.4
## 6 1.71 52 1.67 84
## 7 1.57 61 1.53 52
## 8 1.74 76 1.66 76
## 9 1.73 63 1.59 64
## 10 1.93 71.1 1.93 90

Summary statistics
## # A tibble: 1 x 5
## Count AvHeight SDHeight AvWeight SDWeight
## <int> <dbl> <dbl> <dbl> <dbl>
## 1 20 1.67 0.117 67.2 11.7

1. Calculate β1 and β0 . You can use the short-cuts:

n n n
!2
2 1 X
Xi2
X X
(Xi − X̄) = − Xi
i=1 i=1 n i=1

n n n n
! !
X X 1 X X
(Xi − X̄)(Yi − Ȳ ) = Xi Yi − Xi Yi
i=1 i=1 n i=1 i=1

2. Calculate fits and residuals.

3. Calculate SSE/R, df, MSE/R = σ̂ 2 , RMSE/R = σ̂

4. Calculate SE(βˆ1 ) and SE(βˆ0 ).

5. Calculate t-statistics for each parameter.

6. Conclusion, scope of inference.

2
 1. Calculate β̂1 and β̂0 .

n
1
(Xi − X̄)2 = 57.1214 − 1137.03842 = 0.26948
X

i=1 20

n
X 1
(Xi − X̄)(Yi − Ȳ ) = 2289.782 − (33.72)(1348) = 17.054
i=1 20

17.054
βˆ1 = = 63.28
0.26948

βˆ0 = 67.4 − (63.28)(1.686) = −39.2982

2. Calculate fits and residuals.

## # A tibble: 20 x 4
## X Y Fit Res
## <dbl> <dbl> <dbl> <dbl>
## 1 1.72 67 70.0 -2.98
## 2 1.66 66 66.3 -0.320
## 3 1.59 67 62.0 4.96
## 4 1.5 42 56.5 -14.5
## 5 1.78 83 73.6 9.35
## 6 1.71 52 69.4 -17.4
## 7 1.57 61 60.8 0.177
## 8 1.74 76 71.2 4.79
## 9 1.73 63 70.6 -7.60
## 10 1.93 71.1 82.8 -11.7
## 11 1.75 64 71.8 -7.82
## 12 1.55 65 59.6 5.40
## 13 1.61 75 63.3 11.7
## 14 1.65 70 65.7 4.29
## 15 1.61 55.4 63.3 -7.87
## 16 1.67 84 66.9 17.1
## 17 1.53 52 58.4 -6.38
## 18 1.66 76 66.3 9.68
## 19 1.59 64 62.0 1.96
## 20 1.93 90 82.8 7.19

3
 3. Calculate SSE/R, df, MSE/R = σ̂ 2 , RMSE/R = σ̂

n
res2i = 1895.16
X
SSE =
i=1

df = n − 2 = 20 − 2 = 18

M SE = 1895.16/18 = 105.2867

√
RM SE = 105.2867 = 10.26093

4. Calculate SE(βˆ1 ) and SE(βˆ0 ).

s
1
SE(β1 ) = 10.26093 = 19.76621; df = 37
(20 − 1)(0.1190931)2
v
1 (1.686)2
u
u
SE(β0 ) = 10.26093t + = 33.40472; df = 18
20 (20 − 1)(0.1190931)2

5. Calculate t-statistics for each parameter.

βˆ1 : t = 63.28484/19.76621 = 3.201668; df = 18

βˆ0 : t = −39.29825/33.40472 = −1.1764; df = 18

6. There is strong evidence of a relationship between mass and height for the APES
honours 2024 stats class (slope: two-sided t18 = 3.2; p = 0.004).

Scope of inference.

4
R to make life easier (yes, R can do that sometimes)

library(Sleuth3)
library(tidyverse)

# load data set for analysis

studentData <- read_csv(file = 'StudentData2024.csv')

# fit linear regression

# compare your calculations to those in the model output
model1 <- lm(mass ~ height, data = studentData)
summary(model1)

# create a table with X, Y, fits and residuals

tibble(
X = height,
Y = mass,
Fit = predict(model1),
Res = residuals(model1)
)

# calculate SSR/E, df, MSR/E and RMSR/E from regression output

model1.res <- model1$residuals # or
model1.res <- resid(model1)
SSR <- sum(model1.resˆ2)
df <- model1$df.residual
MSR <- SSR/df
RMSR <- sqrt(MSR) # this value should be the same as 'Residual standard error'
# in the regression output