1 2 1

1

Median 25%-75%

Non-Outlier Range Outliers Extremes

YES NO

RAN 40

60 80 100 120 140 160

PULSE2

RAN

PULSE2

Median 25%-75%

Non-Outlier Range Outliers Extremes SEX: 1

YES NO

40 60 80 100 120 140 160

SEX: 2

YES NO

80 100 120 140 160 180 200 220

WEIGHT 60

62 64 66 68 70 72 74 76

HEIGHT

3

There are 2.5 million gas bottles in Hungary. They want to fill more gas in them. The volume of the bottle limits the maximum ammount of gas. (i.e. the smaller the bottle the higher the pressure in it – for a given ammount of gas)

Gas bottle problem

What is the lot and what is the sample?

Sample Known Histogram

Lot Unknown Density function conclusion

Random variable

5

Continuous random variable

a b x

(

)

∫ ( )

b

a

dx x f b x a P

density function

x

x

F(x)

F(x

)

( ) ( ) ∫ ( )

∞

−

=

≤

= ⁱ

x i

i Px x f xdx

x F

distribution function

Parameter and statistic

• expected value: • sample mean:

• variance • sample variance

∑

=

= ^N

i

xi

x N

1

1

( )

∑

− −

= ^N

i

i x

N x s

1 2 2

1 1

( ) ( )

E x xf x dx

∞

−∞

=

∫

( ) [

]

( )

Var x x E x f x dx

∞

−∞

=

∫

7

The expected value of the volume of gas bottles is 25.8 dm³. What is the expected value in cm³?

( ) ( )

E cx =cE x

( )

( )

Var cx =c Var x

The variance of the volume of gas bottles is 0.25 (dm³)². What is the variance in (cm³)²?

The expected value and variance of the function of random variables

The data in the samples are the results of measurements.

These are subject to error.

Type of measurement errors:

- systematic - random

Why is it important to do independent repetitions?

Independent measurements

Two measurements are independent if their errors are independent.

9

If x₁, x₂, …, x_n are independent and has the same distribution, with E(x) expected value and Var(x) variance.

(

) ( )

E x + + +x x =nE x

The expected value and variance of the function of random variables

(

) ( )

Var x + + +x x =nVar x If x₁, x₂, …, x_nare independent:

(

) ( ) ( )

( )

E x + + +x x =E x +E x + +E x

(

) ( )

( )

( )

Var x+ + +x x =Var x +Var x + +Var x

x

µ

x

f(x)

σ

The most important continuous distribution:

Gauss (normal) distribution

( )









 



 



 −

−

=

2

2 exp 1 2

1

σµ σ

π x x

f Two parameters:

µ

σ

11

Expected value and variance:

( )

µ

E ^Var

( )

^σ

Short notation:

( ^µ

^σ

)

N ^{e.g. N}

( )

Standardisation:

σ µ

= x − z

( )

µ

^σ

( )

( )







−

= exp 2

2

1 z²

z

f π

What is the probability of finding the

x

( µ

σ

µ

σ )

( µ

σ )

( µ

σ )

P

σ µ σ µ

µ x

-1 0 1

µ +σ µ −σ

P(x ≤ µ − σ )

P(x≤ µ + σ )

13

=1

−

= +

σ µ σ µ

upper

z

−1

− =

= −

σ µ σ µ

lower

z

Width of the interval

±

σ

σ

σ

P 0.68268 0.9545 0.9973

The variance of a measurement isσ²=0.25g². The measurement is unbiased. We measure a 10 g etalon weight. In which range will be the measured weight with 95% probability?

(

)

P −z_α < ≤z z_α =

(

)

P

µ −

σ < ≤ +

µ

σ = − α

/ 2 x / 2 0.95

P z_α µ z_α σ

−

 

− < ≤ =

 

 

(

)

P x < ≤x x =

This is the question:

This is what we know from the distribution function:

Connection: x

z

µ

σ

= −

(

)

P − ⋅ < ≤ +x ⋅ =

lower

x xupper

15

The volume of gas bottles are normally distributed with 25.8 dm³ expected value and 0.0625 (dm³)² variance.

What is that minimum volume, that 99.5% of the bottles exceed?

How many percent of the bottles will be in the 25.8±0.3 dm³ interval?

In what interval will be 99% of the bottles?

The variance of a measurement is

σ

1. In which range will be the outcome of the measurement with 99% probability?

2. What is the probability that the measured weight is above 205g?

3. What is the probability that the measured weight is below 200g?

4. What is that maximum value, that the measured weight will not achieve with 90% probability?

Homework

17

The sample mean

(

)

∑

= n xi

x n ...

x n x

x 1 1

2 1

( )

[

( )

]

( )

E 1

( ) ( )

n n

x x Var

Var ^x

x

2

2 σ

σ = = =

Central limit theorem

2 1

( , )

N i i

x N N

µ σ

∑

( , 2 ) x ∼ N

µ σ

The mean of sample elements taken from any

distribution approximately follows Gauss d. around the expected value of the original d. with variance

σ

Sum as well

z x

n

µ σ

= −

19

Calculate the 95% probability interval for the mean of a n = 5 sample taken from a population ofµ=10 andσ²=0.25 !

(

)

P −z_α < ≤z z_α =

(

)

P µ−z_α σ n< ≤ +x µ z_α σ n = −α

/ 2 x / 2 0.95

P z z

α µn α

σ

− < − ≤ =

 

 

(

)

P x < ≤x x =

This is the question:

This is what we know from the distribution function:

Connection: x z

n

µ σ

= −

(

)

P

lower

x xupper

The mean of five measurements is 10, The variance of the measurements is known form previous data:σ²=0.25. In what range can be the true expected value of the measurements with 95%

probability?

(Give a 95% confidence interval for the expected value !)

(

)

P x−z_α σ n < ≤ +µ x z_α σ n = −α

(

)

P − ⋅ < ≤ +µ ⋅ =

The way of thinking is the same as it was on the previous slide, but the inequality is rearranged so, that the expected value (mu) remains in the middle:

21

χ

(chi-square) distribution

∑

= ⁿ

i

zi 1

2

χ2

( )

E χ = = −ν n

0.00 0.05 0.10 0.15 0.20

0 5 10 15 20 25

χ² f(χ²) ν=4

ν=7

ν=10

ν

( )

Var χ = ν = n−

f(χ²⁾

χ² α

χ²_α

23

Distribution of the variance (squared standard deviation) of a sample taken from a normally

distributed population

( )

s n x_i x

i n

2 2

1

1

= −1 −

=

∑

(

)

i n

− = = −

∑

2 2

χ σ ^, ν 1

s²

2 2

= χ σ ν

A sample of 8 elements are taken from a normal distribution having

σ

s

(

)

P χ <χ ≤χ =

2

2 2

s

ν χ σ ⁼

2 2

2 2

upper lower 2

0.95

P χ σ s χ σ

ν ν

 

< ≤ =

 

 

 

(

)

P

s²

2 2

= χ σ ν

2

2 2

lower s 2 upper 0.95

P χ ν χ

σ

 

< ≤ =

 

 

This is the question:

This is what we know from the distribution function:

Connection:

s²_lower s²_upper

25 χ²

f(χ²)

χ²lower χ²upper

0.025

0.025

Critical χ

values

69 .

2 1

lower=

χ

0 .

2 16

upper=

χ

=7

ν

Statistics > Probability Calculator > Distributions...

27

(

)

7 08 . 0 0 . 16 7

08 . 0 69 .

1 2 = < 2≤ =



 



 ⋅ < ≤ ⋅

=P s P s

2 2

2 2

upper lower 2

0.95

P

χ σ

χ σ

ν ν

 

< ≤ =

 

 

 

A sample of 8 elements are taken from a normal distribution. The sample variance is s² = 0.08. Calculate the interval in which the the variance (σ²) is found at 95% probability!

(Give a 95% confidence interval forσ²!)

2 2

2

2 2

lower upper

s s 0.95

P

ν σ ν

χ χ

 

> ≥ =

 

 

 

! Inequality sign is reversed !

Confidence interval

This interval contains the parameter (e.g. m, s…) with 1-α

probability.

Interpreting

If the 95% confidence interval is calculated for the expected value from 100 different sample, than approximately 95 interval contains the true expected value out of the 100.

The confidence interval refers to the PARAMETER.

(Not to x, or s…)

29

Student’s t distribution

t

f(t)

0.0 0.1 0.2 0.3 0.4

-3 -2 -1 0 1 2 3

ν =4

ν =20

ν =1 ^{E t}

( )

t= x- s/ n

µ

n z x

σ µ

= −

Results of 10 measurements:

24.46; 23.93; 25.79; 25.17; 23.82; 25.39; 26.54; 23.85; 24.19;

25.50.

Give the 95% confidence interval for the true (expected) value!

α/2 α/2

-t_α/2 0 t_α/2

f(t)

(

^µ

)

^α

P

t= x- s/ n

µ

(

)

^α

31

F distribution

F s

= s¹

2 1

2

2 2

2 2

/ /

σ

σ σ

σ

2 2

= 2

if

F=s s₁² ₂² two parameters:

ν

ν

33

Critical values for the F distribution

f(F)

F_α F

α

1 2

1 2 1

( , ) 1

( , ) F^α ν ν F _α

− ν ν

=

Two sets of measurements (4 and 7 repetitions) were performed using the same method and device.

Give the 90% probability interval for the ratio of sample variances (squared standard deviations).

The population variances are equal:

σ

σ

(

) (

)

2 2 2 1 0.95 upper

2 2 2 1

lower <s s ≤F P F <s s ≤F =

F P