• Keine Ergebnisse gefunden

Nullhypothese H 0 : X i i.i.d. N (µ 0 , σ 2 ) mit µ 0 = 200, σ = 10

N/A
N/A
Protected

Academic year: 2022

Aktie "Nullhypothese H 0 : X i i.i.d. N (µ 0 , σ 2 ) mit µ 0 = 200, σ = 10"

Copied!
2
0
0

Wird geladen.... (Jetzt Volltext ansehen)

Volltext

(1)

Prof. Peter B¨ uhlmann Mathematik IV: Statistik FS 2013

Musterl¨ osung zu Serie 7

1. a) Test durchf¨ uhren:

Modellannahme: X i : i−te Ammoniumbestimmung. X i i.i.d. N µ, σ 2 mit σ = 10.

Nullhypothese H 0 : X i i.i.d. N (µ 0 , σ 2 ) mit µ 0 = 200, σ = 10

Alternative H A : X i i.i.d. N (µ, σ 2 ) mit µ > 200, σ = 10 (einseitig) Teststatistik: Z = X−µ σ/ n

0

∼ N (0, 1) unter H 0

Verwerfungsbereich: Aus der Normalverteilungstabelle:

K = {z ≥ z 1−α } = [1.64, ∞).

(Dies entspricht dem Verwerfungsbereich [204.1, ∞) auf der Skala von X .) Wert der Teststatistik: z = x−µ σ/ n

0

= 204.2−200 σ/ 16 = 1.68

Testentscheid: 1.68 ∈ K, also wird die Nullhypothese verworfen.

Eine Grenzwert¨ uberschreitung ist statistisch gesichert.

b) Aus a) folgt: Die Nullhypothese kann verworfen werden, falls der Mittelwert aller Messungen gr¨ osser als 204.1 ist,

X > 204.1.

Um die Wahrscheinlichkeit zu berechnen, dass eine Grenzwert¨ uberschreitung nachgewiesen wer- den kann (H 0 verworfen werden kann), geht man wieder zu einer standardisierten normalverteilten Zufallsvariable ¨ uber. Mit µ A = 205 und σ = 10 erh¨ alt man

P[X > 204.1] = P[ X − µ A σ/ √

n > 204.1 − µ A σ/ √

n ]

= P[ X − µ A

σ/ √

n > −0.36]

= P[Z > −0.36]

Dies enspricht also der Wahrscheinlichkeit, dass eine normalverteilte Zufallsvariable Z mit Varianz 1,

Z ∼ N (0, 1),

einen Wert gr¨ osser als −0.36 annimmt. Diese Wahrscheinlichkeit ist wegen der Symmetrie der Normalverteilung gleich zu

P [Z ≤ 0.36] = 0.6406

wie aus der Tabelle entnommen werden kann. Die gesuchte Wahrscheinlichkeit (die Macht des Testes) ist also ungef¨ ahr 64%.

c) Dies ist genau das Niveau des Testes und war als 5% vorgegeben.

(2)

2

2. Die untenstehenden Grafiken zeigen, dass die Form der Verteilung des Mittelwerts von unabh¨ angigen Zufallsvariablen auch dann der Normalverteilung immer ¨ ahnlicher wird, wenn die Variablen selber

¨ uberhaupt nicht normalverteilt sind. An der x-Achse sieht man auch, dass die Varianz immer kleiner wird.

Original

sim

Frequency

0 2 4 6 8 10

0 100 200 300

●●●

●●

●●

●●●

●●

●●●●

●●

●●

●●

●●

●●●

●●●

●●●●●●

●●

●●

●●

●●

●●●●

●●

●●●

●●

●●●

●●●

●●

●●●

●●

●●

●●

●●

●●●

●●●

●●●●

●●

●●

●●

●●●

●●

●●

●●●

●●●

●●●

●●

●●●●

●●

−3 −2 −1 0 1 2 3

0 2 4 6 8 10

Normal Q−Q Plot

Theoretical Quantiles

Sample Quantiles

Mittelwerte von 5 Beobachtungen

sim.mean

Frequency

0 2 4 6 8 10

0 100 200 300

● ●●

● ●

● ●●

● ●

●●

● ●

● ●

●●

●●

● ● ●

● ●

● ●

●●

●●● ●

● ●●

● ●

●●

●●

● ●

● ●●

● ●

●●

● ●

● ●

●●

●●

● ●

● ●

● ●

● ●

● ●

●●

●●●

●●

●●

● ●

●●

●●

● ●●

● ●

●●

● ●

● ●

● ● ●

● ●

● ●●

●●●

● ●●● ●

● ●

● ●

● ●

●● ●

● ●

● ●

● ●

● ●

●●

●●

● ●

●●●●

●●

● ●

● ●

● ●

●●●●

●●

● ●

● ●

● ●

● ● ●●

●●●

●●

●●

● ●

● ●

● ●

● ●

● ●

● ●

●●

●●

●●

● ●

●●●

● ●

● ●

● ●

●●

● ●

●●

●●

● ●

● ●●

●●

● ● ●

● ●

● ●

● ●

● ●

● ●

● ●

●●

●●

●●

●●

● ●

● ● ●

●●

● ●

● ●

●●

●●

● ●●●

● ●

● ●

● ●

● ●

● ●

● ●●

● ●

● ●●

● ●●●

●●

● ●

● ●

● ●

● ●

●●

●●●

● ●

● ●

● ●

●●●

●●

● ●

●●

● ●

● ●

● ●

● ●

●●

● ●

● ● ●

●●

● ●

●●●

● ●●

●●

● ●

●●

● ● ●

●● ●

● ●

● ●

●●

●●

● ●●

●●

● ●

−3 −2 −1 0 1 2 3

0 2 4 6 8 10

Normal Q−Q Plot

Theoretical Quantiles

Sample Quantiles

Mittelwerte von 10 Beobachtungen

sim.mean

Frequency

2 4 6 8 10

0 50 150 250

●●

● ●

●●

●●

● ●

●●

●●

● ●

●●

●●

● ●●●

●●

●●

●●●

● ●●

● ●

●●

● ●

● ●

● ●

● ●

●●

●●

● ●

● ●

●●

●●

● ●●

●●

●●

●●

●● ●

●●

●●

● ●

●●

● ●

●●

●●●

●●

●●

● ●

●●

● ●

● ●

●● ●

●● ●

● ●

● ●

● ●

● ●

●●

● ●

● ●

●●

●●

●●

● ●

● ●●

●●

●●

●●

●●

●●

●●

●●

● ●●

●●●

● ●

●●

●●

●●

● ●

● ●

●●

● ●

●●

● ●

● ●

●●

● ●●

●●

● ●

●●

● ●

●●

●●●

●●

● ●

● ●

● ●●

● ●●

●●●

●●●

● ●

●●

●●

● ●

●●

●●

●●

●●

●●

●●

● ●●

●●

●●

●●

● ●

●●

●●

●●

● ●

●●

● ●

●●

●●

●●

●●

● ●

●●

● ●●

● ●

●●

●●

●●

● ●

●●

● ●

●●

●●●

●●

● ●

●●

●●●

● ●

●●

●●

−3 −2 −1 0 1 2 3

2 4 6 8 10

Normal Q−Q Plot

Theoretical Quantiles

Sample Quantiles

Mittelwerte von 200 Beobachtungen

sim.mean

Frequency

5.5 6.0 6.5 7.0 7.5 8.0

0 50 150 250

●●

●●

●●

●●

●●●

● ●

●●

●●

● ●

●●

●●

●●

●●

●●

●●

●●●

●●

●●

● ●

●●

● ●

●●●

● ●

●●●

●●

●●

●●

● ●

●●

●●

●●

●●

●●

●●

● ●

●●

● ●

●●

● ●

● ●

●●

● ●

●●

●●

●●

●●

● ●

●●

●●

●●

●●

●●

●●

● ●

●●

● ●●

●●

●●

● ●

● ●

●●

● ●

●●

●●

● ●

● ●

● ●

● ●

● ●

●●

●●

●●

● ●

●●

●● ●●

●●

●●

●●●

●●

●●

● ●

● ●

●● ●●

●●

●●●

● ●

●●

●●

●●

● ●●

●●

●●

●●

●●

●●●

●●

●●●

● ●

●●

● ●

●●

● ●●

● ●

● ●

●●

●●

●●

●●●●

●● ●

●●

●●

●●

● ●●●

●●

●●

●●

●●

●●

●●

●●

●●

●●

●●●●

●●

● ●

●●

●●

● ●

●●● ●

●●●●

●●

●●

●●●

●●

●●

● ●

●●●●

●●

●●

●●

● ●

●●

● ●

●●

● ●

●●●●

●●●●

● ●

● ●

● ●

●●

●●

●●

●●

●●

● ●

●●

●●

●●●

●●

●●●

● ●●

●●

●● ●

● ●

●●

● ●

●●

●●

● ●

●●

●●

●●

●●

● ●●

●●

●●

●●

●●

● ●

●●

● ●

● ●

●●●

●●

●●

●●

●●

●●

●●

●●

●●

●●

●●

−3 −2 −1 0 1 2 3

6.0 7.0 8.0

Normal Q−Q Plot

Theoretical Quantiles

Sample Quantiles

Die Grafiken wurden mit folgendem R Code generiert:

> set.seed(1) # fuer Reproduzierbarkeit

> par(mfrow=c(4,2), mar=c(4.1,5.1,1.4,0.5)) # Settings fuer Grafiken

> werte <- c(0,10,11) # moegliche Werte von X

> sim <- sample(werte, 1000, replace=TRUE) # X simulieren

> hist(sim, main = "Original") # Histogramm erstellen

> qqnorm(sim) # Q-Q Plot erstellen

> for (n in c(5, 10, 200)) { # Anzahl Beobachtungen die

# gemittelt werden sim.mean <- rep(0, 1000) # Anzahl Replikate for(i in 1:1000){

sim.mean[i] <-

mean(sample(werte, n, replace=TRUE)) # X simulieren und mitteln }

hist(sim.mean, main =

paste("Mittelwerte von", n, # Histogramm erstellen

"Beobachtungen"))

qqnorm(sim.mean) # Q-Q Plot erstellen

}

Referenzen

ÄHNLICHE DOKUMENTE

−→ Block robuste Regression... statt linearer −→ Glättung, Nichtparametrische Regr... Nicht iterieren!).. 4.8

Einfacher verhält es sich, wenn Eulersche Multiplikato- ren existieren, die nur von einer der beiden Variablen x oder

Universit¨ at T¨ ubingen T¨ ubingen, den 24.01.2018 Mathematisches

Universit¨ at T¨ ubingen T¨ ubingen, den 26.01.2016 Mathematisches

Kann aufgefasst werden als Zuordnung zwischen Formal Algebren → Body-Algebren...

[r]

(Ω, A, µ) sei ein beliebiger Maßraum und ν ein beliebiges

Die Diagonalelemente sind nicht invariant unter einem Phasenfaktor exp ( − iθ) und verursachen daher eine Symmetriebrechung..