Scribd
Upload
35 views

Formulae Corrected

This document provides formulae and procedures for common statistical tests, including: - Two-sample t-test and Wilcoxon rank-sum test for comparing means/medians of independent populations - Paired t-test and Wilcoxon signed-rank test for comparing means/medians of dependent populations - F-test for comparing variances of independent populations - Test for comparing proportions of independent populations - Chi-square test for goodness of fit and test for association - One-way and two-way ANOVA procedures and test statistics.

Uploaded by

Praneeth Bakki
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
35 views

Formulae Corrected

This document provides formulae and procedures for common statistical tests, including: - Two-sample t-test and Wilcoxon rank-sum test for comparing means/medians of independent populations - Paired t-test and Wilcoxon signed-rank test for comparing means/medians of dependent populations - F-test for comparing variances of independent populations - Test for comparing proportions of independent populations - Chi-square test for goodness of fit and test for association - One-way and two-way ANOVA procedures and test statistics.

Uploaded by

Praneeth Bakki
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 3

Some Necessary Formulae

(symbols have their usual interpretation)


Test for equality of means/medians of two independent populations
(a) Two-sample t -test
Test statistic for H0 : x = y
t=

Y H0
X
q
tn1 +n2 2
S n11 + n12

where S 2 =

(n1 1) Sx2 + (n2 1) Sy2


n1 + n2 2

100(1 )% Confidence interval for (x y )


r

1
1

(X Y ) t/2;(n1 +n2 2) S
+ ,
n1 n2

Y ) + t/2;(n +n 2) S
(X
1
2

1
1
+
n1 n2

(b) Wilcoxon rank-sum test


Test statistic for H0 : x = y
W n2 (N + 1)/2 H0
Z=p
N (0, 1)
n1 n2 (N + 1)/12
where W denotes the sum of the ranks received by the second sample, N = n1 + n2 is the
total sample size and the distribution of Z holds approximately for large sample size.
Test for equality of means/medians of two dependent populations
(a) Paired sample t -test
Test statistic for H0 : D = x y = 0

H0
nD
t=
tn1
SD

where

2
SD

1 X
2
=
(Di D)
n 1 i=1

and Di = Xi Yi denotes ith paired differences, i = 1(1)n.


100(1 )% Confidence interval for (x y )


SD
Y ) + t/2;(n1)
(X
n

SD
Y ) t/2;(n1)
(X
,
n

(b) Wilcoxon signed-rank test


Test statistic for H0 : = 0
T + n(n + 1)/4
H0
Z=p
N (0, 1)
n(n + 1)(2n + 1)/24
where T + denotes the sum of the ranks received by the positive differences, denotes
the median of rank differences and n denotes the number of non-zero differences. The
distribution of Z holds approximately for large sample size.
1

Test for equality of variances of two independent populations


(a) F -test
Test statistic for H0 : x = y
H

F = Sx2 /Sy2 0 Fn1 1,n2 1


100(1 )% Confidence interval for x /y


1
F/2;(n1 1),(n2 1)

S2
, F/2;(n21),(n11) x2
Sy

S2
x2
Sy

Test for equality of proportions of two independent populations


Test statistic for H0 : p1 = p2
s


n1 p1 + n2 p2
1 H0
1
N (0, 1) where p =
+
Z = (
p1 p2 )/ p(1 p)
n1 n2
n1 + n2
The distribution of Z holds approximately for large sample sizes.
100(1 )% Confidence interval for p1 p2

s
s

p1 (1 p1 ) p2 (1 p2 )
p1 (1 p1 ) p2 (1 p2 )
(
p1 p2 ) Z/2
+
, (
p1 p2 ) + Z/2
+

n1
n2
n1
n2
Test for goodness of fit
Test statistic for the null hypothesis that the sample frequencies match the
expected frequencies
( I
)
X f2
H
i
T =
n 0 2I1
0
npi
i=1
where fi represents the sample frequency in the ith row, i = 1(1)I, np0i represents the
corresponding expected frequency and n denotes the total sample size.
Test for association
Test statistic for the null hypothesis that the two attributes are independent
"
T =n

I X
J
X
i=1 j=1

#
fij2
H
1 0 2(I1)(J1)
fi0 f0j

where fij denotes the sample frequency of the cell in ith row and jth column, fi0 denotes
ith row total and f0j denotes jth column total, n being the total sample size.

One-way ANOVA
Computational steps
P P i 2
TSS = ki=1 nj=1
xij n x2G , where xG denotes the grand mean
P
SSB = ki=1 (
xi xG )2 ni , where xi denotes the sample mean of the ith group
Pk
SSW = i=1 (ni 1) Si2 , where Si2 denotes sample variance of the ith group
TSS = SSB + SSW
Test statistic
F =

SSB/(k 1) H0
Fk1,nk
SSW/(n k)

under the null hypothesis that the population group means are equal.
Two-way ANOVA
Computational steps
P P
TSS = bi=1 kj=1 x2ij b k x2G
P
x0j xG )2 is the sum of squares due to columns
SSC = b kj=1 (
Pb
xi0 xG )2 is the sum of squares due to rows
SSR = k i=1 (
TSS = SSC + SSR + RSS, RSS being the residual sum of squares
Test statistics
F1 =

SSC/(k 1)
H0
Fk1,(b1)(k1)
RSS/(b 1)(k 1)

under the null hypothesis that the (population) column means are equal.
F2 =

SSR/(b 1)
H0
Fb1,(b1)(k1)
RSS/(b 1)(k 1)

under the null hypothesis that the (population) row/block means are equal.

You might also like