You are on page 1of 1

Hypothesis Testing Roadmap

¾ If P > 0.05, then fail to reject HO Contingency Table


¾ If P < 0.05, then reject HO
¾ Ensure the correct sample size is taken. HO: Data is normal Variable or 2 HO: FA Independent FB
HA: Data is not normal START >>> Attribute Data?
Attribute 1 or 2 Factor?
Factor HA: FA Dependent FB

Levene's Test Stat>Basic Stat>Normality Test or Stat>Tables>Chi2 Test


Stat>Basic Stat>Descriptive Statistics
H O : σ 1 = σ 2 = σ 3 ...
(graphical summary) Variable 1 Factor
H A : σ i ≠ σ j for i ≠ j
(or at least one is different) 1-Proportion Test
Stat>ANOVA>
2 or Data 1 Sample HO: P 1 = P t
Homogeneity of Variance
more
1, 2 or more
not Is data normal?
1 1, 2 or more
1 or >1 Levels? 1 level HA: P 1 ≠ P t
If HO is rejected, then you can levels? Factor Factors? t = target
go no further levels Normal to test
Stat>Basic Stat>
1-Proportion
Fail to reject HO
1 level 2 or more Factors
2-Proportion Test
2 Samples HO: P 1 = P 2
Data Normal DoE
2 levels or Test median
2 levels to test for HA: P 1 ≠ P 2
each 2 levels
> 2 levels? or sigma? Stat>Basic Stat>
2-Proportion
Bartlett's Test
HO: σ1 = σ2 = σ3 ...
2 levels only Test for sigmas
1, 2 or >2 More than HA: σ i ≠ σj for i ≠ j
(or at least one is different) Test for means
Mann-Whitney Test levels? 2 levels
Chi2 Test Stat>ANOVA>Homogeneity of Variance
HO: M 1 = M 2 If sigmas are NOT equal, proceed with caution or use
HO: σ1 = σt Welch's Test, which is not available in Minitab 1-Way ANOVA
HA: M 1 ≠ M 2
HA: σ 1 ≠ σ t (assumes equality of
Stat>Non-parametric> t = target 1 level variances)
Mann-Whitney Test
2 or Medians Stat>Basic HO: μ1 = μ2 = μ3 ...
more
Stat>Display Desc> HA: μ i ≠ μj for i ≠ j
Graphical Summary Test for Test for Is Data (or at least one is different)
levels Mood's Median Test 2 levels
(if target sigma falls mean or sigma? means Dependent?
(used with outliers) Stat>ANOVA>1-Way
between CI, then fail to
H O: M 1 = M 2 = M 3 ... reject HO) (then select stacked or
HA: M i ≠ M j for i ≠ j Test for Test for unstacked data)
(or at least one is different) mean or sigma? means
No,
Stat>Non-parametric> 1-Sample Wilcoxon or Data is drawn Yes,
1-Sample t Test
Mood's test independently Data is
1-Sample Sign HO: μ1 = μt Test for from two Paired
HO: M 1 = M t
Test for HA: μ 1 ≠ μ t sigmas populations
sigmas
HA: M 1 ≠ Mt
t = target
Kruskal-Wallis Test Stat>Basic Stat>
(assumes outliers) t = target
1-Sample t
Stat>Non-parametric> Chi2 Test 2-Sample t Test Paired t Test
HO: M 1 = M 2 = M 3 ...
and either 1-Sample HO: μ1 = μ 2
H A: M i ≠ M j for i ≠ j HO: μ1 = μ2
Sign or 1-Sample HO: σ1 = σt HA: μ 1 ≠ μ2
(or at least one is different) F Test HA: μ1 ≠ μ 2
Wilcoxon HA: σ 1 ≠ σt
Stat>Non-parametric> t = target HO: σ1 = σ2 Stat>Basic Stat> Stat>Basic Stat>
Kruskal-Wallis Stat>Basic HA: σ 1 ≠ σ2 2-Sample t Paired t
Stat>Display Desc> (if sigmas are equal, use
Stat>ANOVA>
Graphical Summary pooled std dev to compare.
Homogeneity of
(if target sigma falls If sigmas are unequal
Variance
between CI, then fail to compare means using
reject HO) unpooled std dev)

You might also like