In: Statistics and Probability
Wind Mountain is an archaeological study area located in southwestern New Mexico. Potsherds are broken pieces of prehistoric Native American clay vessels. One type of painted ceramic vessel is called Mimbres classic black-on-white. At three different sites, the number of such sherds was counted in local dwelling excavations.
site I | Site II | Site III |
61 | 26 | 13 |
34 | 10 | 31 |
28 | 55 | 65 |
18 | 63 | 22 |
78 | 15 | |
50 | 19 | |
26 |
Shall we reject or not reject the claim that there is no difference in population mean Mimbres classic black-on-white sherd counts for the three sites? Use a 10% level of significance.
(b) Find SSTOT, SSBET, and SSW and check that SSTOT = SSBET + SSW. (Use 3 decimal places.)
SSTOT
SSBET |
SSW
Find d.f.BET, d.f.W, MSBET, and MSW. (Use 2 decimal places for MSBET, and MSW.)
dfBET
dfW
MSBET |
MSW |
Find the value of the sample F statistic. (Use 2 decimal places.)
What are the degrees of freedom?
( numerator)
(denominator)
(c) Find the P-value of the sample test statistic. (Use 4
decimal places.)
(f) Make a summary table for your ANOVA test.
Source of Variation |
Sum of Squares |
Degrees of Freedom |
MS | F Ratio |
P Value | Test Decision |
Between groups | . | |||||
Within groups | ||||||
Total |
A ) level of significance is 10% ( 0.10)
Hypothesis
No all means are equal
B) Enter data in excel sheet like this
Go to DATA ribbon. Choose ANOVA : One factor from data Analysis menu.
Output from Excel( ANOVA Table )
SSTOT = 726.765
SSBET = 722.4076
SSW = 6541.357
There are k = 3 treatment and N = 17 total observations therefore,
dfBET = ( k - 1 ) = ( 3 - 1) = 2
dfW = ( N - k ) = ( 17 - 3 ) = 14
MSBET = SBET / dfBET = 722.4076 / 2 = 361.2038
MSW = SW / dfW = 6541.357 / 14 = 467.2398
F = MSBET / MSW = 0.770359
The degree of freedom for F statistics is
dfN = 2
dfD = 14
C ) The p value for the test can be obtained using Excel functon FDIST(F,dfN,dfD) as :
p value = FDIST(0.773059,2,14) = 0.480333
Since p value is greter than level of significance at 0.10 , We do not reject null hypothesis
At 10% level of significance there is sufficient evidence that the means are equal