Non-Parametric Test
(Distribution-free Tests)
• Refers to the interference procedures that do not require the population
distribution to be normal or some other form speci ed in terms of
parameters.
WILCOXON RANK SUM TEST
• Popular because they apply to a very wide variety of population distribution.
• They utilize simple aspects of sample data, such as measurements order
relationships or category frequencies.
• Can be determined without regard to the shape of the underlying population
distribution.
What type of observations are To conduct non-parametric test, follow the five-step approach:
especially suited to non-parametric Set up hypotheses and select the level of signi cance (α). Analogous to parametric testing, the research
hypothesis can be one- or two- sided (one- or two-tailed), depending on the research question of interest.
analysis? Select the appropriate test statistic. The test statistic is a single number that summarizes the sample
information. In nonparametric tests, the observed data is converted into ranks and then the ranks are
summarized into a test statistic.
Characteristics like degree of apathy, taste preference, surface gloss and etc. cannot be
Set up decision rule. The decision rule is a statement that tells under what circumstances to reject the null
evaluated on an objective numerical scale and an assignment of numbers is, therefore, bound hypothesis. Note that in some nonparametric tests we reject H0 if the test statistic is large, while in others we
to be arbitrary. reject H0 if the test statistic is small. We make the distinction as we describe the di erent tests.
For example, when people are ask to express their views on a 5-point rating scale the numbers Compute the test statistic. Here we compute the test statistic by summarizing the ranks into the test statistic
identi ed in Step 2.
have little physical meaning beyond the fact that higher scores indicate greater agreement.
Conclusion. The nal conclusion is made by comparing the test statistic (which is a summary of the
Strongly Disagree Indifferent Agree Strongly information observed in the sample) to the decision rule. The nal conclusion is either to reject the null
Disagree Agree hypothesis (because it is very unlikely to observe the sample data if the null hypothesis is true) or not to reject
1 2 3 4 5 the null hypothesis (because the sample data are not very unlikely if the null hypothesis is true).
fi
fi
fi
fi
Let’s recall the assumptions of the two-sample Assumption no. 1
t test for comparing two population means: (Dependent variable should be measured at the ordinal or continuous level)
Consider a clinical trial where study participants are asked to
Dependent variable should be measured at the ordinal or continuous level rate their symptom severity following 6 weeks on the assigned
treatment. Symptom severity might be measured on a 5 point
Independent variable should consist of two categorical, "related ordinal scale with response options: Symptoms got much
groups" or "matched pairs“
worse, slightly worse, no change, slightly improved, or much
improved. Suppose there are a total of n=20 participants in
The distribution of the di erences between the two related groups needs to the trial, randomized to an experimental treatment or placebo,
be symmetrical in shape. and the outcome data are distributed as shown in the gure.
The distribution of the outcome (symptom severity) does not
appear to be normal as more participants report improvement
in symptoms as opposed to worsening of symptoms.
Assumption no. 2 Assumption no. 3
(Independent variable should consist of two categorical, "related (The distribution of the differences between the two related groups needs to
groups" or "matched pairs“) be symmetrical in shape.)
For example, you might have measured 10 individuals' performance in a spelling Samples of observations from each of two populations A and B
test (the dependent variable) before and after they underwent a new form of containing nA and nB observations respectively. We wish to test the
computerized teaching method to improve spelling. hypothesis that the distribution of X-measurements in population A is the
same as that in B, which we will write symbolically as H0 : A = B.
You would like to know if the computer training improved their spelling
performance. The first related group consists of the subjects at the beginning
(prior to) the computerized spelling training and the second related group consists
of the same subjects, but now at the end of the computerized training.
The Wilcoxon signed-rank test can also be used to compare different subjects
within a "matched-pairs" study design, but this does not happen very often.
ff
fi
What is Wilcoxon Test?
Wilcoxon Rank A popular non-parametric test used to compare
outcomes between two independent groups.
Also called Mann Whitney U test or Mann Whitney
Sum Test Wilcoxon test or Wilcoxon rank sum test.
Used to test whether two samples are likely to derived
from he same population.
What is Wilcoxon Test? Example:
•Two geological formations are compared with respect to richness of mineral
content. The mineral contents of 7 specimens of ore collected from formation 1
In contrast, the null and two-sided research hypotheses for and 5 specimens collected from formation 2 are measured by chemical analysis.
the non-parametric test are: The following data are obtained:
H0: The two populations are equal versus.
H1: The two populations are not equal.
The procedure for the test involves pooling the
Do the data provide strong evidence that formation 1 has a higher
observations from the two samples into one combined mineral content than formation 2? Test with a near .05.
sample, keeping track of which sample each observations
comes from, and then ranking lowest to highest from 1 to
n1 + n2, respectively.
13
Hypotheses Rank Sum Statistics
Ho: n formation 1 = n formation 2 n(n+1)/2 12(12+1)/2= 78
HA: n formation 1 ≠ n formation 2
Critical Values
The PROBLEM claims that Near 0.05
•Formation 1 > Formation 2
Assigning Ranks
Formation 1 7.6 11.1 6.8 9.8 4.9 6.1 15.1 Rank
Sum
Rank 9 11 8 10 5 6 12 61
Formation 2 4.7 6.4 4.1 3.7 3.9
T2= 17 Do Not T1= 61
Rank 4 7 3 1 2 17 Reject Ho
Reject Ho Reject Ho
Total 78 22 43
Test Statistics
Conclusion
T1= 61 T2=17 Total= 78 The observed value fall in the rejection
region so the Null hypothesis is rejected at
α=0.053
