High | Medium | Low | Total | |
---|---|---|---|---|
Peak | 20 (15) | 7 (9) | 3 (6) | 30 |
Trough | 30 (50) | 40 (30) | 30 (20) | 100 |
Rising | 20 (15) | 8 (9) | 2 (6) | 30 |
Falling | 30 (20) | 5 (12) | 5 (8) | 40 |
Total | 100 | 60 | 40 | 200 |
H1: Some other distribution.
Following the hint we must determine 5 equ-probable intervals
for a normal distribution and see if indeed all the observations fall
rather evenly into these intervals.
For any distribution the 20, 40,60,and 80th percentiles divide the
distribution into 5 equally probable intervals. The z-values for these
percentiles are -0.84, -0.25, 0.25, and 0.84 (from the z-table or e.g.
NORMSINV(.20). The corresponding percentiles for a normal distribution
with mean = 14 and standard deviation = 5 are resp:
Range | fo | fe | stat |
---|---|---|---|
9.8 or less | 9 | 8 | 0.125 |
9.8-12.75 | 8 | 8 | 0.000 |
12.75-15.25 | 6 | 8 | 0.500 |
15.25-18.20 | 5 | 8 | 1.125 |
18.20 or more | 12 | 8 | 2.000 |
Alternatively we could have defined other ranges, say by dividing the entire range of values (roughly from 1 to 21) equally into 5 categories of width 4 resulting in intervals: 1-5; 5-9; 9-13; 13-17; 17-21 and calculated the Chi-square on the basis of the observed and the expected frequencies falling into these intervals.
Alternative: with equally spaced intervals
Note: In calculating the Cum_P use either e.g.,
=NORMDIST(5,14,5,1) which returns the probablity of the random variable
to assume a value up to 5.
Or if you want to use the tables; calculate the z -value (e.g. (5-14)/5
= -1.8 ) and look up the probability for z=1.8 which is 0
.4641. This is the probablity from 0 to 1.8 (the blue shaded
area)
or, due to the symmetry of normal distribution, to the left, from
0 to -1.8 . Therefore the probability up to z = -1.8 is 0.5 -
0.4641= 0.0359.
[go back] [Any
questions?]
Ho: Sample comes from a normal distribution (with some mean
and some standard distribution)
H1: From some other distribution
Sales | z(UL) |
Cum Prob |
Int Probability | Observed(Expected) | Stat |
65.5 or less |
-1.3889 |
.0824 |
.0824 | 10 (16.49) | 2.552 |
66.5 - 70.5 | -0.8333 |
.2023 |
.1199 | 20 (23.98) | 0.660 |
70.5 - 75.5 | -0.2778 |
.3906 |
.1883 | 40 (37.65) | 0.146 |
75.5 - 80.5 | 0.2778 |
.6094 |
.2188 | 50 (43.76) | 0.889 |
80.5 - 85.5 | 0.8333 |
.7977 |
.1883 | 40 (37.65) | 0.146 |
85.5 or more | infinity |
1.000 |
.2023 | 40 (40.47) | 0.005 |
[go back] [Any
questions?]
Additional Problems:
Range | fo | z | Cum Prob | Int Prob | fe | |
---|---|---|---|---|---|---|
<2.6 | 6 | -1.6 | .0548 | .0548 | 8.22 | .5995 |
2.6 - 3.79 | 30 | -.8 | .2119 | .1571 | 23.59 | 1.7613 |
3.8 - 4.99 | 41 | .0 | .5000 | .2881 | 43.22 | .1142 |
5.0 - 6.19 | 52 | .8 | .7881 | .2881 | 43.22 | 1.7829 |
6.2 - 7.39 | 12 | 1.6 | .9452 | .1571 | 23.59 | 5.5709 |
7.40 - and up | 9 | Inf | 1.0000 | .0548 | 8.22 | .0740 |
Total | 150 | 1.0000 | 150 | 10.0028 |
x | fo | P(x) | fe | Stat |
---|---|---|---|---|
0 | 22 | .04979 | 24.89 | 0.33633393 |
1 | 74 | .14936 | 74.68 | 0.00620268 |
2 | 115 | .22404 | 112.02 | 0.07922641 |
3 | 95 | .22404 | 112.02 | 2.58622415 |
4 | 94 | .16803 | 84.02 | 1.18652484 |
5 | 80 | .10082 | 51.41 | 17.3698376 |
6 or more | 20 | .08392 | 41.96 | 11.492 |
500 | 500 | 33.0564 |
Critical value of Chi-square at .05 with 6 degrees of freedom = 12.592. Since the observed Chi-square exceeds this, the null hypothesis that the sample came from a poisson distribution with mean 3 is rejected.