Professional Documents
Culture Documents
ECN 416
Non-Parametric Test of Hypothesis
INTRODUCTION
• Most of the Test of hypothesis procedures discussed
earlier are based on the assumption that the
random samples are selected from a normal
population. Traditionally, these testing procedures
are known as Parametric Methods as they depend
on the parameters, viz., mean or proportion or
standard deviation etc.
• These parametric tests have used the parametric
statistic of sample that comes from the population
being tested.
INTRODUCTION Contd.
• The need for techniques that apply more broadly
has led to the development of Non-Parametric
Methods.
• Non-Parametric Methods do not require that the
underlying population be normal – or indeed
that they have any single mathematical form and
some even apply to non-numerical data.
• Non-parametric Methods are also referred to as
distribution-free methods.
ADVANTAGES OF NON-PARAMETRIC
METHODS
• Non-parametric Methods are generally simple
to understand, quicker and easier to apply
when the sample size id small.
• They do not require lengthy and laborious
calculations and hence they are less time
consuming.
• They do not require the assumptions that a
population is distributed in the shape of a
normal or another specified shape.
ADVANTAGES OF NON-PARAMETRIC
METHODS Contd.
• They can be applies to all types of data –
qualitative (nominal scaling) data in rank form
(ordinal scaling) as well as data that been
measured more precisely (internal or ratio scaling)
• They have inherently greater range of applicability
because of milder assumptions.
• Non-parametric tests make fewer and less
stringent assumptions (that are more easily met)
than the parametric test.
DISADVANTAGES OF NON-PARAMETRIC METHODS
• When data are only roughly scaled; for example when only
comparative rather than absolute magnitudes are available.
As with clinical data, perhaps patients can only be classified
as better, unchanged or worse. ( largest, second largest,
……….., smallest)
RANK SUM TESTS
• Rank sum test are a whole family of tests. We
shall concentrate only on the following two
members of this family:
• Mann-Whitney U Test
• H0 : µ1 = µ2 against H0 : µ1 ǂ µ2
without assuming whether the population
samples have roughly the shape of a normal
distribution.
Mann-Whitney U Test Procedure.
• Step 1 Set the null and the alternative
hypothesis: H0: µ1 = µ2 against H0 : µ1 ǂ µ2
Mean: µ U = n 1 n2 / 2
Bran
dA 2.1 4.0 6.3 5.4 4.8 3.7 6.1 3.3
Bran
dB 4.1 0.6 3.1 2.5 4.0 6.2 1.6 2.2 1.9 5.4
EXAMPLE Contd.
• Test the hypothesis, at 0.05 the level of
significance, that the average caffeine
contents of the two brands are equal against
the alternative that they are unequal.
SOLUTION
Ori 0.6 1.6 1.9 2.1 2.2 2.5 3.1 3.3 3.7 4.0 4.0 4.1 4.8 5.4 5.4 6.1 6.2 6.3
gin
al
Da
ta:
Ra 1 2 3 4 5 6 7 8 9 10. 10. 12 13 14. 14. 16 17 18
nk 5 5 5 5
SOLUTION Contd.
• The ranks of the observation belonging to the
smaller samples are in bold form.
• R1 = 4+8+9+10.5+13+14.5+16+18 = 93
• R2 = 1+2+3+5+6+7+10.5+12+14.5+17 = 78
• n1 = 8, n2 = 10
SOLUTION Contd.
• U-Statistic:
U = n1 n2 +[ n1 (n1 +1)] / 2 - R1
= 8 x 10 +[(8x9) / 2] – 93 = 23
• Mean of U = µu = n1 n2 / 2 = 10x8 / 2 = 40
• Variance of U = σu2
= [(n1 n2) (n1 + n2 + 1)] / 12
= [10 x 8 (10+8+1)] / 12 = 380 / 3 = 126.67
• σu = √126.67 = 11.25
SOLUTION Contd.
• Here n2 = 10, so we can use the statistic
Z = [U - µu] / σu = [23 – 40] / 11.25
= -17 / 11.25 = -1.51
Therefore:|Z| = 1.51
• The table value of Zα = 1.96
• Since 1.51 < 1.96, we cannot reject the null
hypothesis. We therefore conclude that there is no
significant difference in the average caffeine contents
of the two brands of cigarettes
Kruskal-Wallis Test or H-Test
Method I 78 62 71 58 73
Method II 76 85 77 90 87
Method III 74 79 60 75 80
SOLUTION
Ma 58 60 71 62 73 74 75 76 77 78 79 80 85 87 90
rks
Ran 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
ks
SOLUTION Contd.
Method I 10 3 4 1 5 R1 = 23
Method II 8 13 9 15 14 R2 =59
Method III 6 11 2 7 12 R3 = 38
SOLUTION Contd.
• H = 12 / n(n+1)[ ∑1k Ri2 / ni] - 3(n+1)
= 12 / 15x16[ R12 / [ R12 / n 1 + R22 / n 2 +
R32/ n 3 ] – 3 x 16
= 12 / 15x16[ R12 / [ 23 / 5 + 59/ 5 + 38 / 5 ]
– 3 x 16
= 1 /5 [529 + 3481 = 1444] - 48
= 5454 / 100 - 48 = 6.54
SOLUTION Contd.
The degrees of freedom = k – 1 = 3 -1 = 2
Therefore
χ2 (2, 0.05) = 5.991
Decision: Reject H0 if H > χ2 (2, 0.05)
Now 6.54 > 5.991
Therefore the null hypothesis is rejected and the
alternative hypothesis H1 is accepted. We conclude
that there is no difference in the three teaching
methods.
SPEARMEN RANK CORRELATION COEFFICIENT