Inferential Statistics Part C
Hypothesis tests
Eva Riccomagno, Maria Piera Rogantin
DIMA – Universit`a di Genova
http://www.dima.unige.it/~rogantin/UnigeStat/
Part C
Hypothesis tests on the mean of a population
0. Review
1. Aside of probability. Normal (or Gaussian) random variable 2. Test on the mean of a Normal variable – known variance
(running example: toxic algae)
(a) Composite hypotheses (b) The power function
0. Review
A hypothesis testing is formed by - a null and alternative hypothesis
- a test statistics (function of the sample variables) - a rejection region/decision rule
Significance level/Critical value/Test statistic: all contribute to the definition of the rejection rule
Type I and II errors
Example: if R0 = {x such that T (x) > s} then:
α = P(T > s | H0) reject H0 when it’s true β = P(T ≤ s | H1) retain H0 when it’s false
Formulation of the hypotheses and “form” of R0
Example. The quantity of tomatoes in a can is nominally set at 100 g.
Formulate a statistical hypothesis test as if you were:
1. the Federal Trade Commission 2. an unscrupulous shareholder
3. the worker in charge of the quality control of the can tomato filling machine.
Indicate to which hypotheses the following R0 correspond:
R0 = {x s.t. T (x) > s}
R0 = {x s.t. T (x) < s1} ∪ {x s.t. T (x) > s2} R0 = {x s.t. T (x) < s}
In case of two-sided test s1 and s2 are such that
α1 = P(T < s1 | H0) α2 = P(T > s2 | H0) with α1 + α2 = α If the test statistic has a symmetrical distribution often α1 and α2 are set as: α1 = α2 = α/2
Consider Ex. 2 of Assignment 5. X ∼ B(20, p)
What happens when the hypotheses change?
1. H0 : p = 0.3 H1 : p = 0.5 α = 0.05 one-sided right 2. H0 : p = 0.5 H1 : p = 0.3 α = 0.05 one-sided left
> p_0=0.3;p_1=0.5;
> s05_right=qbinom(1-0.05,20,p_0);s05_right [1] 9 ### same result as p_1=0.7
> s05_left=qbinom(0.05,20,p_1)-1;s05_left ### note -1 [1] 5
One-sided right R0 = {x > 9}; one-sided left R0 = {x < 5}.
Notice that for x = 5, 6, 7, 8, 9 the decision of the two tests is different (small size and p0 “close” to p1)
What happens when α change?
1. H0 : p = 0.3 H1 : p = 0.5 α = 0.05 3. H0 : p = 0.3 H1 : p = 0.5 α = 0.01
> s01_right=qbinom(1-0.01,20,p_0);s01_right [1] 11
R0 with α = 0.01 is smaller than R0 with α = 0.05
1. Aside of Probability. Normal (or Gaussian) random variable
Probability density functions and cumulative distribution functions
µ = −1,0, 5 σ = 1
−4 −2 0 2 4 6 8
0.00.20.4
Normal −− sigma = 1
−4 −2 0 2 4 6 8
0.00.20.4
Normal −− sigma = 1
−4 −2 0 2 4 6 8
0.00.20.4
Normal −− sigma = 1
−4 −2 0 2 4 6 8
0.00.40.8
Normal −− sigma = 1
mu= − 1 mu = 0 mu = 5
−4 −2 0 2 4 6 8
0.00.40.8
Normal −− sigma = 1
mu= − 1 mu = 0 mu = 5
−4 −2 0 2 4 6 8
0.00.40.8
Normal −− sigma = 1
mu= − 1 mu = 0 mu = 5
µ = 0
σ = 0.5,1, 3
−5 0 5
0.00.20.40.60.8
Normal −− mu = 0
−5 0 5
0.00.20.40.60.8
Normal −− mu = 0
−5 0 5
0.00.20.40.60.8
Normal −− mu = 0
−5 0 5
0.00.40.8
Normal −− mu = 0
sigma=0.5 sigma=1 sigma=3
−5 0 5
0.00.40.8
Normal −− mu = 0
sigma=0.5 sigma=1 sigma=3
−5 0 5
0.00.40.8
Normal −− mu = 0
sigma=0.5 sigma=1 sigma=3
The density function of X ∼ N (µ, σ2) is: f (x) = √ 1
2π σ2 exp
−(x−µ)
2
2σ2
Some properties of the Normal random variables
• For X ∼ N (µ, σ2) let Y = aX + b. Then Y ∼ N (aµ + b, a2σ2) In particular
Z = X − µ
σ ∼ N (0, 1) Z is called standard Normal variable
• The sum of Normal variables is a Normal variable.
In particular for X1, . . . , Xn independent and identically dis- tributed (i.i.d.) with Xi ∼ N (µ, σ2) for all i = 1, . . . , n, then
X ∼ N µ, σ2 n
!
2. Test on the mean of a Normal random variable with known variance
2. (a) Composite hypotheses
Running example:
X models the concentration of toxic alga blooms
The statistical model is:
X ∼ N (µ, σ2) with σ2 known
and experts set a bathing alert if µ > 10000 cells/liter. Thus H0 : µ ≥ 10000 H1 : µ < 10000
The significance level of the test is set at α = 5%
If we reject H0 we can swim with 5% probability of side effects due to the alga
Test statistics X
Sample size: n = 10 σ = 2100 cells/liter Set α = 5%
The test hypotheses are both composite
First we consider the two cases 1. H0 : µ=10000 H1 : µ= 8500 2. H0 : µ=10000 H1 : µ<10000 and next
3. H0 : µ≥10000 H1 : µ<10000
10000
Case 1.
Assume H0: X ∼ N (10000, 21002/10)
R0 = {x < s} where s is such that PX < s|µ = 10000 = α Get s = 8908 with R
mu0=10000;std=2100/sqrt(10) c1= qnorm(.05,mu0,std);c1
If the test statistic value on the sample is less than 8908, then we reject H0 Spot the probability of type I error in the plot.
If x is larger than 8908, then the prob- ability of type II error is
β = P X > s|µ = 8500 Get β = 27% with R
mu1=8500;1-pnorm(c1,mu1,std)
We retain H0 with a large probability of type II error
10000
8500 retain H
reject H0 0
10000
8500 retain H
reject H0 0
Case 2.
H0 : µ=10000 H1 : µ<10000 R0 does not change
as it is computed under H0
The probability β of II type error be- comes a function of µ as it is com-
puted under H1 namely µ < 10000 reject H 10000retain H
0 0
Case 3.
H0 : µ≥10000 H1 : µ<10000
Keeping the same R0, the prob- ability of type I error, denoted by α(µ), is smaller than α
α(µ) < α
The probability β of the type II error is a function of µ under H1, as in
10000 retain H
reject H0 0
2. (b) The power function P(θ) of a test
Consider a generic test on a parameter θ:
H0 : θ ∈ Θ0 H1 : θ ∈ Θ1
Running example. H0 : µ ≥ 10000 H1 : µ < 10000 Then Θ0 = [10000, +∞), Θ1 = (−∞, 10000)
The power function of a test is the probability to reject H0 as a function of the parameter θ
P (θ) = P(T ∈ R0|θ) =
( 1 − β(θ) if θ ∈ Θ1 correct decision α(θ) if θ ∈ Θ0 type I error
Note that α(θ) ≤ α
1. the critical value is the smallest s s.t. P (T > s|H0) < α 2. the critical region is the largest R0 s.t. P (T ∈ R0|H0) < α
Power function:
P (θ) = P (T ∈ R0|θ) =
( 1 − β(θ) if θ ∈ Θ1 correct decision α(θ) if θ ∈ Θ0 type I error
Running example:
R0 = {X < 8908}
P (µ) = PX < 8908 | µ ∈ R P (10000) = 0.05 = α
0 1
α
10000 8500
1-β(8500)
std=2100/sqrt(10);c1=qnorm(.05,10000,std) mu=seq(7000,11000)
p=pnorm(c1,mu,std)
plot(mu,p,type="l",lwd=3, col="red")
Power and sample size
The probability to reject H0, when it is false, grows as the sam- ple size grows
The probabilities of the two type of errors decrease as the sample size grows
Running example
H0 : µ ≥ 10000 and H1 : µ < 10000 R0 = {Xn < 8908}
P (µ) = P Xn < 8908 | µ ∈ R
n = 10 red n = 20 blue 0
1
α
10000 0
1
α
10000
9500 Θ
Θ1 0
If the values of the parameter under H0 and under H1 are “close”
(e.g. 10000 and 9500 respectively), only with large sample the probability of correct decision is large
The power for one-sided and two-sided tests
One-sided: H0 : µ ≥ 10000 and H1 : µ < 10000
R0 = {X < 8908} P (µ) = PX < 8908| µ ∈ R
Two-sided: H0 : µ = 10000 and H1 : µ 6= 10000 R0 = {X < 8700} ∪ {X > 11300}
P (µ) = PX < 8700 | µ ∈ R + PX > 11300) | µ ∈ R
One-sided red
H0 : µ ≥ 10000 H1 : µ < 10000
Two-sided blue
H0 : µ = 10000 H1 : µ 6= 10000
0 1
α
10000
mu=seq(7000,13000); mu0=10000; std=2100/sqrt(10) c1_u=qnorm(.05,mu0,std); p=pnorm(c1_u,mu,std)
c1_b=qnorm(.025,mu0,std);c2_b=qnorm(.975,mu0,std);p_b=pnorm(c1_b,mu,std)+1-pnorm(c2_b,mu,std) limx=c(7000,13000);limy=c(0,1)
plot(mu,p,type="l",lwd=3,xaxt="n",yaxt="n",xlab=" ",ylab=" ",col="red",xlim=limx,ylim=limy) par(new=T)
plot(mu,p_b,type="l",lwd=3,xaxt="n",yaxt="n",xlab=" ",ylab=" ",col="blue",xlim=limx,ylim=limy) axis(2, at = 0,0, las=1,cex=1.2,col="blue");axis(2, at = 1,1, las=1,cex=1.2,col="blue")
axis(2, at =0.05,expression(alpha), las=1,cex=1.2,col="blue") axis(1, at = mu0,mu0, las=1,cex=1.2,col="blue")
abline(h=c(0,1));abline(h=0.05,v=c(mu0),lty=3,lwd=2);abline(h=0.05,lty=3,lwd=2)