MULTIPLECOMPARISON PROCEDURE FOR USE WITH FRIEDMAN TEST

Learning outcome: Students are able to calculate and use multiple-comparison formula to solve problems.

For treatment test (Friedman test) their data just allow them to conclude that not all sampled populations or all treatment effects are identical. So we use multiple-comparison procedure to see where the differences located. ( ONLY USE WHEN WE REJECT H0 )

FORMULA : | Rj Rj | z

Where; Rj and Rj is the jth and jth treatment rank total where z is a value from table A.2 that corresponding to

WAY TO USE FRIEDMAN MULTIPLE-COMPARISON

STEP 1 : MAKE SURE THAT ITS REJECT H0, THEN WE CAN USE MULTIPLE-COMPARISON STEP 2 : CALCULATE A.2 STEP 3 : FIND VALUE Z STEP 4 : COMPARED | Rj Rj | WITH Z TO FIND VALUE Z BY USING TABLE

STEP 5 : ONLY | Rj Rj | Z DIFFERENCES.

WILL RESULT THE PAIR

EXAMP LE

FROM PREVIOUS EXAMPLE (GROUP 12 FRIEDMAN TEST)

Example 1 (Page 265) Hall et al. * compared three methods of determining serum amylase values in patients with pancreatitis. The result are shown in table 7.2. We wish to know whether these data indicate a difference among the three methods. Given

=00 .0
* Hall, F.F., T. W. Culp,T. Hayakawa, C. R. Ratliff, and N. C.
Hightower,"An Improved Amylase Assay Using a New Starch Derivative, Amer. J. Clin. Pathol.,53 (1970),627-634

Serum amylase values (enzyme units per 100 ml of serum) in patients with pancreatitis
Table 7.2
Methods of determination Specimen
1 2 3 4 5 6 7 8 9 A 4000 1600 1600 1200 840 352 224 200 184 B 3210 1040 647 570 445 156 155 99 70 C 6120 2410 2210 2060 1400 249 224 208 227

Hypotheses
Ho : MA = MB = MC H1 : At least one equality is violated (claim)

Test statistic
b = 9, k = 3

After convert the original observations to ranks, we have

Methods of determination Specimen
1 2 3 4 5 6 7 8 9 A 2 2 2 2 2 3 2.5 2 2 RA = 19.5 RB = 9 B 1 1 1 1 1 1 1 1 1 C 3 3 3 3 3 2 2.5 3 3 RC = 25.5

By equation, we have k = 0 b =0 ,

0 0 k r0 = 0R bk ( k + 0 j = )

0 j

0( k 0 + b )

0 0 0 = (11 + 0 + .1) (1 1 1 .1111 0 )( )( ) (0 0 0 )( )( )

= 11 11 1 .0 1 = 0.0 0

DECISION From table A.11,

(0 , k0 = 00, k . 0 0 ) (000) = 0 0 .00 0 ,0 .
=0

0. > 00 0 00 . 0
Since

then we reject

H0

CONCLUSION Enough evidence to support the claim that the three methods does not all yield identical results.

Since we reject H0 , then we can find where the differences are located by using Multiple-comparison Procedure.
From above we know that : b=9 k=3 = 0.05 Ra = 19.5 Rb = 9 Rc = 25.5 THEN, WE KNOW THE FORMULA IS : | Rj Rj| z

AND Z IS A VALUE FROM TABLE A.2 CORRESPONDING TO FIRST, WE FIND = 0.05 3(3-1) = 0.0083

Then find z value: Z = 0.5 0.00833 = 0.4916667 approximate to 0.4917 (table A.2) z = 2.39 Then we substitute value z into : z We get z = 2.39 = 10.1399 Then : | Rj Rj| 10.1399

Now we want to check where the differences occur. So the three pairs of differences | Rj Rj| are | Ra Rb| = | 19.5 9 | = 10.5 | Ra Rc| = | 19.5 25.5 | = 6 | Rb Rc| = | 9 25.5 | = 16.5 Thus, we can conclude that, pairs that yield different result is pair Ra Rb and Rb Rc . Because it is greater and equal to 10.1399.

EXERCISE