DSC 211
CLASS NOTES 21, page 1
Comparing Two Populations
REVIEW
Target Population: e.g., Customer Incomes (X)
MOD 1:
Questions (1 or 2sided) about:
across k categories
Analysis: 6 Step Hypothesis Testing
"Conclusion About
Specific Question (one or twosided)
the Claim"
1
Ho & Ha > Ha is same as the question
IN SHORT:
2
Test Stat
Is sample
Sample evidence inconclusive. Cannot conclude (Ha)
3
Alpha = usually 0.05
inconsistent
4
Reject Rule > Use TINV, CHIINV, NORMINV
with Ho ?
Reject rule is consistent with Ha.
If so, then
5
Calculate the Test Stat (from the sample results)
"yes" to
NOW:
Part 2 of our DSC 211 Course
pvalue > Use TDIST, CHIDIST, NORMDIST
bus question.
Comparing 2 or more Populations
6
Reject Ho or Do not reject Ho
Issue:
Population 1
Data:
Assembly
Plant 1 Assembly Times
Plant 2 Assembly Times
Times at
2 Plants
BQues:
Question (another way to state it):
Do the sample results provide evidence that the averages are different?
1
2
because we know its distribution
(2) If
Ho is true: t distribution
with df we can calculate
with df =
3
alpha = .05
4
Reject Ho if t < (low neg value) or t > (high pos value)
2sided Hyp
Use =TINV(.05,df)
5
from the two sample results
calculate these 1st four values
df =
use formula above to calc df
t=
use formula above to calc t
pvalue=
use TDIST with the t and df just calculated
Sample (n)
> Test Statistic
:
tTest for Mean, t
= (xbar  mu
o
) / (s/n^.5)
μ
 average income & some value, mu
o
ChiSq Test for Variance, ChiSq
= (n1)*s^2/var
o
σ
2
or var
 variability of incomes & a value, var
o
Normal Test for Prop, pbar
with stdev = (p
o
*(1p
o
)/n)^.5
p
 proportion of incomes > some value, p
o
ChiSq Goodness of Fit Test, ChiSq
= Sum of (fe)^2/e
p1,p2,p3, .
..
 distribution of incomes
First  comparing means (
μ
's)
xbar
1
xbar
2
Sample from 1 (n
1
)
s
1
^2
Sample from 2 (n
2
)
s
2
^2
μ
1
 ave time
μ
2
 ave time
Common question, "Are the assembly times different?" Which usually means >
Are
μ
1
and
2
different?
(Or, Do
μ
1
and
μ
2
differ by more or less than D?)
(Ques could be Is
μ
1
greater than
μ
2
?)
Ho:
μ
1
=
μ
2
,
(or,
μ
1

μ
2
= D, where
D=0
in this case)
Ha:
μ
1
<>
μ
2
,
Test Statistic. (1)
t = [(xbar
1
xbar
2
)  0]
(
(s
1
2
/n
1
+ s
2
2
/n
2
)^0.5
(s
1
^2/n
1
+ s
2
^2/n
2
)^2
(s
1
^2/n
1
)^2/(n
1
1) + (s
2
^2/n
2
)^2/(n
2
1)
xbar
1
=
xbar
2
=
s
1
^2=
s
2
^2=
This preview has intentionally blurred sections. Sign up to view the full version.
View Full Documentcan (cannot) conclude . . .
6
Reject or Do Not Reject Ho.
NOTE:
We will do calculations manually with the following example.
However, in practice we will use excel's tool:
Example:
Issue:
Two possible methods of assembly for new product.
Two Sample t Test> Unequal Variances
Question:
From data gathered using the two methods, can we conclude that the avg assembly times differ?
[If so, how?]
Data:
Samples of 25 randomly selected workers; recorded assembly times for each.
May not be explicit.
6.29
min
6.02
min
Diff =
0.27
What do you think about bus question
1. Hypotheses:
just looking at xbars  before Analysis?
This is the end of the preview. Sign up
to
access the rest of the document.
 Spring '07
 Dunne
 Variance

Click to edit the document details