Chapter 15

Lot-by-Lot Acceptance Sampling for Attributes

LEARNING OBJECTIVES
After completing this chapter you should be able to:
1. Understand the role of acceptance sampling in modern quality control systems
2. Understand the advantages and disadvantages of sampling
3. Understand the difference between attributes and variables sampling plans, and the major types of
acceptance-sampling procedures
4. Know how single-, double-, and sequential-sampling plans are used
5. Understand the importance of random sampling
6. Know how to determine the OC curve for a single-sampling plan for attributes
7. Understand the effects of the sampling plan parameters on sampling plan performance
8. Know how to design single-sampling, double-sampling, and sequential sampling plans for attributes
9. Know how rectifying inspection is used
10. Understand the structure and use of MIL STD 105E and its civilian counterpart plans
11. Understand the structure and use of the Dodge-Romig system of sampling plans

IMPORTANT TERMS AND CONCEPTS

100% inspection Lot tolerance percent defective (LTPD)
Acceptable quality level (AQL) LTPD plans
Acceptance-sampling plan MIL STD 105E
ANSI/ASQC Z1.4, ISO 2859 Multiple-sampling plan
AOQL plans Normal, tightened, and reduced inspection
Attributes data Operating-characteristic (OC) curve
Average outgoing quality Random sampling
Average outgoing quality limit Rectifying inspection
Average sample number curve Sample size code letters
Average total inspection Sequential-sampling plan
Dodge-Romig sampling plans Single-sampling plan
Double-sampling plan Switching rules in MIL STD 105E
Ideal OC curve Type-A and Type-B OC Curves
Lot disposition actions Variables data
Lot sentencing
15-2 CHAPTER 15 LOT-BY-LOT ACCEPTANCE SAMPLING FOR ATTRIBUTES

EXERCISES

Note: Many of the exercises in this chapter are easily solved with spreadsheet application software. The
BINOMDIST, HYPGEOMDIST, and graphing functions in Microsoft Excel were used for these solutions.
Solutions are in the Excel workbook Chap15.xls.

15.1. 
An accounting firm uses sampling methods in its client auditing processes. Accounts of a particular type are
grouped together in a batch size of 25. The auditor is concerned about erroneous accounts escaping the
auditing process. Sampling and auditing the accounts is time consuming and very expensive, and a random
sample of size n = 5 is about the largest sample that can practically be used. Suppose that the batch of accounts
contains one erroneous account. What is the probability that the sample that is selected contains the erroneous
account?

The sampling distribution for this problem is the hypergeometric distribution, with N = 25, n = 5, D = 1. Looking
for the probability that x = 1.

In Minitab,

Probability Density Function

Hypergeometric with N = 25, M = 1, and n = 5

x P( X = x )
1 0.2

Alternatively, in Excel

Probability that the sample contains the erroneous account is 0.20.

 CHAPTER 15 LOT-BY-LOT ACCEPTANCE SAMPLING FOR ATTRIBUTES 15-3

15.2. 
Reconsider the situation described in Exercise 15.1. Suppose that the batch of accounts contains two erroneous
accounts. What is the probability that the random sample of size n = 5 that is selected contains at least one of
the two erroneous accounts?

The sampling distribution for this problem is the hypergeometric distribution, with N = 25, n = 5, D = 2. Looking
for the probability that x = 1 or 2.

In Minitab,

Probability Density Function

Hypergeometric with N = 25, M = 2, and n = 5

x P( X = x )
0 0.633333
1 0.333333
2 0.033333
3 0.000000
4 0.000000
5 0.000000

Alternatively in Excel,

Probability that the sample contains the erroneous account is 0.36.

15-4 CHAPTER 15 LOT-BY-LOT ACCEPTANCE SAMPLING FOR ATTRIBUTES

15.3. 
Reconsider the situation described in Exercise 15.1. How many erroneous accounts must be in the batch of
accounts for a random sample of size n = 5 to have a probability of at least 0.50 containing the erroneous
account?

The sampling distribution for this problem is the hypergeometric distribution, with N = 25, n = 5, D varies.
Looking for the D that gives Pr{x > 0} = 0.5:

If there are D = 3 erroneous accounts in the population of 25 accounts, there is a probability of 0.504 that a
sample of n = 5 contains at least one of the erroneous accounts.
 CHAPTER 15 LOT-BY-LOT ACCEPTANCE SAMPLING FOR ATTRIBUTES 15-5

15.4. 
Hospital personnel routinely examine patient records for error, such as incomplete insurance information, on
incomplete patient history, or missing/incomplete medical records. On average, about 250 new patients are
admitted each day. Historically, about 5% of these records have contained errors. If a random sample of 50 new
patient records is checked each day, what is the probability that this sample will contain at least one patient
record with missing information?

The sampling distribution for this problem is the hypergeometric distribution, with N = 250, n = 50,
D = 250  5% = 12.5. Looking for the probability that x  1.

In Minitab, for 12 erroneous records in the population:

Probability Density Function

Hypergeometric with N = 250, M = 12, and n = 25

x P( X = x )
0 0.273995

Or Pr{x  1} = 1 – 0.27 = 0.73

And for 13:

Probability Density Function

Hypergeometric with N = 250, M = 13, and n = 25

x P( X = x )
0 0.245214

Or Pr{x  1} = 1 – 0.25 = 0.75

The probability that a sample of size 50 contains at least one patient record with missing information is about
0.74.

Consider whether the binomial approximation to the hypergeometric may be used. The ratio of n / N = 50 / 250
= 0.2, which exceeds the suggested 0.1. For comparison though, a binomial with p = 0.05 and n = 50:

Probability Density Function

Binomial with n = 50 and p = 0.05

x P( X = x )
0 0.0769450

Or Pr{x  1} = 1 – 0.08 = 0.92, which differs significantly from the exact solution.
15-6 CHAPTER 15 LOT-BY-LOT ACCEPTANCE SAMPLING FOR ATTRIBUTES

15.5.
Draw the type-B OC curve for the single-sampling plan n = 50, c = 1.
 CHAPTER 15 LOT-BY-LOT ACCEPTANCE SAMPLING FOR ATTRIBUTES 15-7

15.6.
Draw the type-B OC curve for the single-sampling plan n = 100, c = 2.
15-24 CHAPTER 15 LOT-BY-LOT ACCEPTANCE SAMPLING FOR ATTRIBUTES

15.19.
Consider rectifying inspection for single sampling. Develop an AOQ equation assuming that all defective items
are removed but not replaced with good ones.

AOQ  Pa  p  (N  n) N  Pa  (np)  (1  Pa )  (Np)

15.20.
A supplier ships a component in lots of size N = 3,000. The AQL has been established for this product at 1%. Find
the normal, tightened, and reduced single-sampling plans for this situation from MIL STD 105E, assuming that
general inspection level II is appropriate.

N = 3000, AQL = 1%
General level II
Sample size code letter = K
Normal sampling plan: n = 125, Ac = 3, Re = 4
Tightened sampling plan: n = 125, Ac = 2, Re = 3
Reduced sampling plan: n = 50, Ac = 1, Re = 4

15.21.
Repeat Exercise 15.20, using general inspection level I. Discuss the differences in the various sampling plans.

N = 3000, AQL = 1%
General level I
Normal sampling plan: Sample size code letter = H, n = 50, Ac = 1, Re = 2
Tightened sampling plan: Sample size code letter = J, n = 80, Ac = 1, Re = 2
Reduced sampling plan: Sample size code letter = H, n = 20, Ac = 0, Re = 2

15.22.
A product is supplied in lots of size N = 10,000. The AQL has been specified at 0.10%. Find the normal,
tightened, and reduced single-sampling plans from MIL STD 105E, assuming general inspection level II.

N = 10,000; AQL = 0.10%; General inspection level II; Sample size code letter = L
Normal: up to letter K, n = 125, Ac = 0, Re = 1
Tightened: n = 200, Ac = 0, Re = 1
Reduced: up to letter K, n = 50, Ac = 0, Re = 1
 CHAPTER 15 LOT-BY-LOT ACCEPTANCE SAMPLING FOR ATTRIBUTES 15-25

15.23.
MIL STD 105E is being used to inspect incoming lots of size N = 5,000. Single sampling, general inspection level II,
and an AQL of 0.65% are being used.

(a) Find the normal, tightened, and reduced inspection plans.

N = 5000, AQL = 0.65%; General level II; Sample size code letter = L
Normal sampling plan: n = 200, Ac = 3, Re = 4
Tightened sampling plan: n = 200, Ac = 2, Re = 3
Reduced sampling plan: n = 80, Ac = 1, Re = 4

(b) Draw the OC curves of the normal, tightened, and reduced inspection plans on the same graph.
15-26 CHAPTER 15 LOT-BY-LOT ACCEPTANCE SAMPLING FOR ATTRIBUTES

15.23.(b) continued
 CHAPTER 15 LOT-BY-LOT ACCEPTANCE SAMPLING FOR ATTRIBUTES 15-27

15.24.
A product is shipped in lots of size N = 2,000. Find a Dodge–Romig single-sampling plan for which the LTPD = 1%,
assuming that the process average is 0.25% defective. Draw the OC curve and the ATI curve for this plan. What
is the AOQL for this sampling plan?

The AOQL is 0.21%.

Note that this solution uses the cumulative binomial distribution in a spreadsheet formulation. A more precise
solution would use the hypergeometric distribution to represent this sampling plan of n = 490 from N = 2000,
without replacement.
15-28 CHAPTER 15 LOT-BY-LOT ACCEPTANCE SAMPLING FOR ATTRIBUTES

15.25.
We wish to find a single-sampling plan for a situation where lots are shipped from a supplier. The supplier’s
process operates at a fallout level of 0.50% defective. We want the AOQL from the inspection activity to be 3%.

(a) Find the appropriate Dodge–Romig plan.

Dodge-Romig single sampling, AOQL = 3%, average process fallout = p = 0.50% defective
Minimum sampling plan that meets the quality requirements is 50,001  N  100,000; n = 65; c = 3.

(b) Draw the OC curve and the ATI curve for this plan. How much inspection will be necessary, on the average,
if the supplier’s process operates close to the average fallout level?

let N  50,001
Pa  Binom(3,65,0.005)  0.99967
ATI  n  (1  Pa )(N  n)  65  (1  0.99967)(50,001  65)  82

On average, if the vendor’s process operates close to process average, the average inspection required will be
82 units.

(c) What is the LTPD protection for this plan?

LTPD = 10.3%
 CHAPTER 15 LOT-BY-LOT ACCEPTANCE SAMPLING FOR ATTRIBUTES 15-29

15.26.
A supplier ships a product in lots of size N = 8,000. We wish to have an AOQL of 3%, and we are going to use
single sampling. We do not know the supplier’s process fallout but suspect that it is at most 1% defective.

(a) Find the appropriate Dodge–Romig plan.

N = 8000; AOQL = 3%; p  1%

n = 65; c = 3; LTPD = 10.3%

(b) Find the ATI for this plan, assuming that incoming lots are 1% defective.

c 3
Pa   binomial(n, p)   b(65,0.01)  0.9958
d 0 d 0

ATI  n  (1  Pa )(N  n)  65  (1  0.9958)(8000  65)  98

(c) Suppose that our estimate of the supplier’s process average is incorrect and that it is really 0.25% defective.
What sampling plan should we have used? What reduction in ATI would have been realized if we had used the
correct plan?

N = 8000; AOQL = 3%; p  0.25%

n = 46; c = 2; LTPD = 11.6%

c 2
Pa   binomial(n, p)   b(46,0.0025)  0.9998
d 0 d 0

ATI  n  (1  Pa )(N  n)  46  (1  0.9998)(8000  46)  48