Chapter 15 Athena en lecture Flashcards
central question of hypothesis testing
how extreme is a sample if the null hypothesis is true?
test statistic
a standardized measure of how far the point estimate falls from the parameter value if the null hypothesis is true
sampling distribution
a probability distribution that specifies probabilities for the possible values the statistic can take
central limit theorem
for any large enough sample, the sampling distribution is approximately normally distributed with
mean = u
stdev = o/wortel(n)
wilcoxon test assumpties
- data are ranked ordered
- two independent samples
- (no assumption about distribution)
wat is H0 bij wilcoxon test
H0=the population distributions of the quantitative score are identical
wat is Ha bij wilcoxon test
Ha= the population distributions of the quantitative score differ in such a way that the expected average ranks are also different
test statistic bij wilcoxon test
= difference in average of all ranks in that group/condition
wat is de p bij wilcoxon test
altijd two sided! dus x2
ranks berekenen
alles in een grote kolom
dan ranks invullen (GOED TELLEN: kom je uit op totale n? en alle ties goed?)
dan gemiddelde van alle ranks terwijl ze in de rij staan!!
dan terug zetten in kolommen
dan Ri berekenen per groep
hoe de pvalue berekenen bij kruskal wallis test
1-chi.dist
met df= g-1
wanner sign test gebruiken
bij two dependent groups (als niet normaal verdeeld is)
wat is het principe van de sign test
+ en - bij pretest post test!
posttest-pretest
wat als het 0 is bij de sign test
? invullen
assumpties van sign test
random samples of matched pairs, unequal values for each pair
wat is belangrijk bij de sign test qua hypothese?
als het one sided is: goed kijken of de hypothese is pretest>post test of andersom!
dan op basis daarvan de formule doen. + = GOEDE UITKOMST
wat is H0 bij sign test
H0= the samples contain independent pairs, p op + = 0,5
wat is de test statistic bij sign test
hoeveel mensen een + hadden
wat voor soort variabele bij non-parametric tests
ordinal!!
wanneer nonparametric tests gebruiken
We do not know whether the data are normally distributed (e.g.,
because of small sample size)
The data are not normally distributed and sample size is too low for
the central limit theorem to apply
waar wordt elk van de 3 testen voor gebruikt
- wilcoxon: two independent groups
- kruskal wallis: more than two independent groups
- sign test: two dependent groups
wilcoxon test stappen
When conducting Wilcoxon test:
Convert to ranks (remember the Excel tip!)
Calculate the mean rank per group
Calculate the difference between mean ranks
Locate the difference in the sampling distribution
Small n→ look at the sample space, then construct sampling distribution
(construction of sampling distribution not needed for exam)
Large n→ normal approximation
kruskal wallis test stappen
When conducting Kruskal-Wallis test:
Convert to ranks (remember the Excel tip!)
Calculate the appropriate statistics (overall mean rank, mean rank per
group)
Fill in the formula for the test statistic
Locate the test statistic in the Chi Square distribution (df = #groups -1)
sign test stappen
When conducting sign test:
Compute the difference scores
Count the number of positive difference scores
Locate the test statistic in the sampling distribution
For n ≥ 30 → normal approximation
For smaller n→ binomial distribution
parametric testing
Parametric
* More assumptions
* Sensitive to outliers
* Preferred when much data
* More power
* Also applicable to more
complex models
nonparametric testing
Non-parametric
* Less assumptions
* Less sensitive to outliers
* Preferred when data is limited
* Less power
* Not as well-developed for more
complex models
waar zijn rank analyses tegen bestemd
outliers!
hoe p value berekenen bij de wilcoxon test
aflezen uit sampling distribution.
difference in mean ranks gebriuken: kijken naar deze value of meer extreem.
stel difference = 2, dan P= 2* (P(2)+P(3)+P(4)… etc.
of z score gebruiken (zie formule blad)
GOED KIJKEN NAAR OF HET ONE SIDED OF TWO SIDED IS!! ANDERS P VALUE ALTIJD *2!!!!!
OKE
