One-sided alternative
0
0: =
H µ µ H1:µ ≠ µ0
two-sided
0 0:
H µ≥µ H1:µ <µ0 one-sided
Example 3
Ni content of a solution is measured: 3.25, 3.27, 3.24, 3.26, 3.24
The required minimum value is 3.25 g/cm3 Is it reached?
1
0 0:
H µ≥µ H1:µ <µ0
t
n s n s x n s
t0=x−µ0 = −µ+µ−µ0
(
µ <µ0)
<0 if H1
lower tail, lower limit
(minimum value is not reached)
0 0:
H µ≥µ H1:µ < µ0
The hypothesisedµ0=3.25 value is above the lower confidence limit (p>0.05), accepted:
failed to reject ...
not proved that at most...
0 0
H :µ µ≤ H :1 µ µ> 0
The hypothesisedµ0=3.25 value is below the upper confidence limit (p>0.05), accepted:
failed to reject ...
not proved that at most...
If the null hypothesis is accepted it does not „prove”
Comparing two variances ( F test)
2 2 2 1 1 :σ >σ H
2 2 2 1 0:
H σ =σ
The test statistic: 2;
(
1 1 2 1)
2 2 1
0 = n − ,n −
s F s
In case of one-sided alternative:
Fα
s s12 / 22 >
The null hypothesis is rejected if
5
In case of two-sided alternative: H1:σ12 ≠σ22
The null hypothesis is rejected if
2 2 1
2 2 1
F -a/
s
s < 2 2
2 2 1
Fa/
s s >
or
1 / 22
2
1 s ≥
s it is sufficient to check the upper limit
using
Two-sample t test
Two independent samples 1 2
2 2 2 1 2
1,n ; s ,s ; x,x n
Assuming the equality of variance for the two populations (to be checked through F-test):
( )
0:
H0 Ex1−x2 =µ1−µ2 =
2 2 2
1 σ
σ =
7
( ) ( )
[
1 1]
2 1
2 2 2 1
2 1 2 1
2 = s n − +s n −
- +n s n
( )
2 1
2 1 2 1
1 1
n s n
x x E x t=x
+
−
−
−
2 2
1+ −
=n n ν
( )
0:
H0 E x1−x2 =µ1−µ2 =
The test statistic:
,
(
1−1) (
+ 2 −1)
= n n
ν
σ
1σ
2 2
= 2
The assumption is checked through
F
test2 1
2 1
0 1 1
n s n
x
= x t
+
−
Example 4
(Box-Hunter-Hunter: Statistics for Experimenters, J. Wiley, 1978, p. 97)
The wear of two kinds of raw material is compared as shoe soles on the foot of 10-10 boys (shoes1.xls).
Is the difference of means and variances significant at 0.05 level?
n mean sample variance
A 10 10.63 6.009
B 10 11.04 6.343
TABLE 4.3. Data on the amount of wear measured with two different materials A and B, boy’s shoes example*
boy material A material B B – A
difference d
1 13.2(L) 14.0(R) 0.8
2 8.2(L) 8.8(R) 0.6
3 10.9(R) 11.2(L) 0.3
4 14.3(L) 14.2(R) -0.1
5 10.7(R) 11.8(L) 1.1
6 6.6(L) 6.4(R) -0.2
7 9.5(L) 9.8(R) 0.3
8 10.8(L) 11.3(R) 0.5
9 8.8(R) 9.3(L) 0.5
10 13.3(L) 13.6(R) 0.3
average difference 0.41
Example 15
(Box-Hunter-Hunter: Statistics for Experimenters, J. Wiley, 1978, p. 97)
wear
material A material B boys
5 7 9 11 13 15
0 1 2 3 4 5 6 7 8 9 10 11
Paired t test ( ) ( )
0:
H0 E xi =E yi =
i i
i x y
d = −
( ) ( ) ( )
di E xi E yiE = −
: H0
n d
d i
∑
i=
( )
1
2
2
n- d - d
s i
i d
∑
= s / n
= d t
d 0
one-sample
t
test for the differences13
Paired t test ( )
00:E d =
H di =xi −yi
1. 1.
2. 2.
dependent samples
boys
B - A
-0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0 1.2
0 1 2 3 4 5 6 7 8 9 10 11
149 .
2 =0
sd sd = 0.149 =0.386 0.122
10 386 .
0 =
n = sd
4 . 122 3 . 0
41 . 0
0 = =
t
beta
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0
two-sample paired
OC curve for the boys shoes example
Statistics > Basic Statistics/Tables t-test, dependent samples
Box Plot of multiple variables wear 5v*10c
Median; Box: 25%-75%; Whisker: Non-Outlier Range
Median 25%-75%
Non-Outlier Range
A B DIFF
-2 0 2 4 6 8 10 12 14 16
Testing goodness of fit
It is to be judged if the data may come from a certain distribution
Normality test
graphical tests statistical tests
Open:
body.sta
Statistics > Distribution fitting > continuous Distributions > Normal
Variable: WEIGHT, Distribution: Normal Kolmogorov-Smirnov d = 0,04966, p < 0,20, Lilliefors p < 0,01
Chi-Square test = 15,64733, df = 7 (adjusted) , p = 0,02854
140 145 150 155 160 165 170 175 180 185 190 195 200 205
Category (upper limits) 0
10 20 30 40 50 60 70 80 90 100 110
No. of observations
Graphs > 2D Graphs > Normal Probability Plots…
Normal Probability Plot of WEIGHT (body 4v*500c)
-4 -3 -2 -1 0 1 2 3 4
Expected Normal Value
Large samples:
Kolmogorov–Smirnov test
The data are grouped into classes, at least 5 classes are required.
χ2 -test
The data are grouped into classes, at least 5 occurrences are required in a class.
Smaller samples:
Shapiro–Wilk test
Statistical tests for goodness of fit
Statistics > Distribution fitting > continuous Distributions > Normal
Options tab
Kolmogorov-Smirnov test: yes(continuous)
Variable: WEIGHT, Distribution: Normal
Kolmogorov-Smirnov d = 0,04966, p < 0,20, Lilliefors p < 0,01 Chi-Square test = 15,64733, df = 7 (adjusted) , p = 0,02854
30 40 50 60 70 80 90 100 110
No. of observations
Graphs > 2D Graphs > Normal Prbability Plots…
Shapiro–Wilk test
Normal Probability Plot of WEIGHT (body 4v*500c)
140 150 160 170 180 190 200 210
Obs erved V alue -4
-3 -2 -1 0 1 2 3 4
Expected Normal Value
WEIGHT: SW-W = 0,9927; p = 0,0160
Normal Probability Plot of maleWEIGHT (body unstack 4v *253c)
-1 0 1 2 3 4
pected Normal Value
Variable: maleWEIGHT, Distribution: Normal
Kolmogorov-Smirnov d = 0,07996, p < 0,10, Lilliefors p < 0,01 Chi-Square test = 4,56330, df = 4 (adjusted) , p = 0,33511
150 155 160 165 170 175 180 185 190 195 200 205
Category (upper limits) 0
10 20 30 40 50 60 70 80 90
No. of observations