- •It is expressed in the same unit of measure as the observed variable
- •It is the point in a distribution of measurements about which the sum of deviations are equal to zero .
- •It does not explain the variability inside the range !
- •Very simple and straightforward measure of dispersion
- •It is the sum of the squared deviations from the mean divided by n
- •Variance of the sample mean → σ2/n
- •4 Prezent
- •Is a statement to be tested
- •It is always a statement of no difference, always contains a statement of equality
- •Impact of the factor on quantitative output
- •Impact of 1 factor – one way anova
- •Impact of 2 factors – two way anova
- •Independency of samples
- •In case of h1
- •Is the deviation random or not (error)
- •10 Prez
Test
Impact of the factor on quantitative output
Impact of 1 factor – one way anova
Impact of 2 factors – two way anova
…
Etc.
Analysis of Variance - a set of methods used to test the impact of one or more factors on population mean
an inferential method that is used to test the equality of three or more population means
One-way ANOVA - Decomposition of total variance on sub-variances
Sub-variances contain information which influence xij values in observed distribution
s2 → variance due to model s12 + residual variance sr2
One-way ANOVA model
One-way ANOVA model
balanced model (equal n)
unbalanced model (different n)
model presumptions
Normal distribution of all samples
Independency of samples
Homogeneity of variances
Null hypothesis statement
H0: m1 = m2 = … = mm
H0: a1 = a2 = … = am = 0
Alternative hypothesis H1
at least one of the population means is different from the others
Multiple comparison
In case of h1
Scheffé´s method
Tuckey´s method
Duncan´s method
Kramer´s method
What is null hypothesis of ANOVA?
What are assumptions of ANOVA?
Why we make detailed pairwise comparison after we reject null hypothesis? What methods can we use?
What represents residual variance?
What is difference between balanced and non-balanced ANOVA model?
What is difference between one-way and two-way ANOVA?
Parametric tests
Parameters testing
Assume normal distribution (F-test, t-test)
Nonparametric tests
Distribution is not required
Quantitative and qualitative variables
Simple calculation (especially for small samples)
Lower power
Two sample Wilcoxon test
Nonparametric equivalent of the two sample t-test
Test of two independent samples
To test the hypothesis that two independent samples X=(x1, x2,…,xm) and Y=(y1, y2,…,yn) are from the same population
procedure
all values are ranked in ascending order (pooled sample) with ties assigned the average of the next available ranks
Wilcoxon test
Nonparametric equivalent of the paired t-test
We assess whether two dependent samples are from the same population
Procedure
For each pair of dependent observations (xi, yi) compute difference di=xi-yi
all values are ranked in ascending order (pooled sample) with ties assigned the average of the next available ranks
Absolute values of differences are ranked in ascending order (zero differences are not considered)
Summarize separately order of positive and negative differences
W+ = sum of order for positive differences
W- = sum of order for negative differences
Test criterion is to lower value from W+ a W-
W=(W+ ,W- )
H0 is rejected, if W≤Wα,n, where
Wα,n critical table value
n…number of nonzero differences
Sign test
For two dependent samples
Procedure
The test criterion Z represents the lower number from positive and negative differences
Z=min (Z+, Z-)
if Z < Za; n, H0 is rejected
n - number of nonzero differences
a - significance level
Kruskall – Wallis test
Nonparametric equivalent of one-way ANOVA
Test of the hypothesis that m independent samples with sizes n1, n2,…nm are from the same distribution
Procedure
all values are ranked in ascending order (pooled sample) with ties assigned the average of the next available ranks
Dixon test of extrem deviations
