0% found this document useful (0 votes)
31 views36 pages

Control Charts and Hypothesis Testing

Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
31 views36 pages

Control Charts and Hypothesis Testing

Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd

Hypothesis Testing

Shilpa G.

05-09 Feb 2024

1 / 36
Control Charts
1. (3-sigma) x̄ chart

1.1 Process mean and standard deviation known


1.2 Process mean and standard deviation unknown
2. R chart
3. p chart
4. np chart
2 / 36
x̄ Chart: Process Mean and Standard Deviation Known

UCL = µ + 3σx̄
LCL = µ − 3σx̄

3 / 36
x̄ Chart: Process Mean and Standard Deviation Unknown
• The estimator µ̂ for µ is x where

x̄1 + x̄2 + · · · + x̄k


x=
k
where

x̄j = mean of the j-th sample, j = 1, 2, . . . , k


k = number of samples

• For finding estimate for σ, we could compute the sample


deviation for each sample and by pooling the resulting
estimate. This process is a bit cumbersome.
• We use sample ranges for this purpose. Recall that for a RS
X1 , X2 , . . . , Xn , Range= X(n) − X(1) .

4 / 36
• For normal random variables it can be shown that the ratio of
the expected value of the sample range R to the standard
deviation σ that depends only on the sample size.
E [R]
d2 = .
σ

• we use R̄ average of range which is defined as


R1 + R2 + · · · + Rk
R̄ =
k
where
Rj = range of the j-th sample, j = 1, 2, . . . , k
k = number of samples
This gives us

σ̂ = .
d2
5 / 36
6 / 36

UCL = x + 3 √
d2 n

LCL = x − 33 √
d2 n

7 / 36
8 / 36
R Chart
m
P
Rj
j=1
In this case, µ̂R = R̄ = and it can be shown that when
m
sampling from a normal distribution, a good estimator for σR is
d3 R̄
σ̂R = .
d2
This gives us control limits
d3
UCL = R̄ + 3σ̂R = R̄D4 with D4 = 1 + 3
d2
d3
LCL = R̄ − 3σ̂R = R̄D3 with D3 = 1 − 3
d2
Using previous data, calculate R. chart.

9 / 36
p Chart (Proportion Defective)

• Π= population proportion of defective items


• X = number of defective items in a sample of size n
X
• p= sample proportion of defective items
n P
pi
• Successive estimate for Π by p1 , p2 , . . . , pm is p̄ =
m
• µp = Π and σp2 = Π(1 − Π)/n
• Suppose we have m samples, each of size n then
r
p̄(1 − p̄)
µ̂p = p̄ and σ̄p =
n
• whenever nΠ ≥ 5 and n(1 − Π) ≥ 5, we have

10 / 36
p Chart and np Chart

For q
a p chart For np chart
p̄ + 3 p̄(1− p̄) p
UCL np̄ + 3 np̄(1 − p̄)
q n
p̄ − 3 p̄(1− p̄) p
LCL n np̄ − 3 np̄(1 − p̄)

Example
An electronics firm produces computer memory chips. Statistical
quality control methods are to be used to monitor the quality of
the chips produced. A chip is classified as defective if any flaw is
found that will make the chip acceptable to the buyer. To set up a
p chart to monitor the process, 300 chips are sampled on each of
20 consecutive days. Write control limits.

11 / 36
HomeWork: §19.3 (T1). 12 / 36
Difference between two population means

13 / 36
So far, we saw hypothesis testing for mean, median
of single population. Before looking for hypothesis
testing for variance, we see how to proceed for two
populations.

In general, we work with more than one populations


simultaneously. Many times, we can generalize
methods that we use for single population.
Following is one such example. We will have a look
at it and then come back to single population.

14 / 36
σ1 and σ2 known
If both populations have a normal distribution, or if the sample
sizes are large enough
• The 100(1 − α) CI of µ1 − µ2 is given by
s
σ12 σ22
(x̄1 − x̄2 ) ± zα/2 +
n1 n2
where 1 − α is the confidence coefficient.
• Hypothesis test for µ1 − µ2

15 / 36
16 / 36
Inferences about Population Variance

θ = σ2
X (xi − x̄)2
θ̂ = s 2 =
n−1
For finding 100(1 − α)% confidence interval, we need to find a, b
such that
P[a ≤ σ 2 ≤ b] = 1 − α.

17 / 36
Hypothesis Testing:

18 / 36
Summary

19 / 36
20 / 36
Goodness of fit

So far, we saw hypothesis testing of mean, variance, median


for a single population. There is one more test, called
Goodness of Fit Tests where we check whether population
comes from normal population using given random sample.

Goodness of Fit Test is special case of Chi-Square Test. Other


tests are test for independence and comparing proportions. We
will study all these after discussing tests for two populations.

21 / 36
(From the book of Ross)

22 / 36
Two populations: means with σ1 and σ2 unknown

This is called Smith-Satterthwaite degree of freedom.

23 / 36
24 / 36
FYI

25 / 36
26 / 36
27 / 36
Matched samples

• A random sample is chosen which is used to get sample values


for both the populations that are in consideration.
• For example,

• In the matched sample design data about populations are


gathered under similar conditions hence this design often leads
to a smaller sampling error than the independent sample
design.
We use the following terms and steps.

28 / 36
• d = a reminder that the matched sample provides difference
data.
• µd = the mean of the difference in values for the population 1
and 2. For samples x1 , x2 , . . . , xn and y1 , y2 , . . . , yn of size n,
we have di = xi − yi for all i.
s
¯2
P P
di (di − d)
d¯ = , sd = .
n n−1

• If we know that the population of differences has a normal


distribution or sample size is large, we use the test statistics

d¯ − µd
t= √ ∼ Tn−1 .
sd / n

29 / 36
Remark

A matched sample procedure for inferences about


two population means generally provides better
precision than the independent sample approach;
therefore it is the recommended design. However, in
some applications the matching cannot be achieved,
or perhaps the time and cost associated with
matching are excessive. In such cases, the
independent sample design should be used.

30 / 36
31 / 36
Two Populations with σ12 = σ22

32 / 36
F Distribution
A random variable X is said to have F distribution with parameters
d1 and d2 is
U1 /d1
X =
U2 /d2
where Ui ∼ χ2 (di ) and U1 , U2 are independent.

33 / 36
34 / 36
35 / 36
36 / 36

You might also like