Bayesian Special Type Double Sampling Plan with Two
Decision Criteria for Discrete Prior Distribution
[Link] MENON
University Research Scholar
Department of Statistics
Bharathiar University
Coimbatore-641 046
TamilNadu, INDIA
E-mail Address –[Link]@[Link]
[Link]
Associate Professor and Head
Department of Statistics
Bharathiar University
Coimbatore-641 046
TamilNadu, INDIA
E-mail Address-osdeepamenon@[Link]
ABSTRACT
This paper presents Bayesian Sampling Plan based on Special Typ e of Double Sampling
(STDS) Plan. The selection of parameters for STDS plan through Bayes solution considering
double binomial as prior distribution is presented. Illustration is provided for easy selection of plan
parameters.
INTRODUCTION:
To evaluate the minimum expected value for desirable lesser cost function, it is
appropriate to consider some prior distribution for the quality characteristic or lot quality.
Calvin (1984) provides procedures and tables for implementing Bayesian Sampling
Plans. A set of tables presented by Oliver and Springer (1972) are based on the assumption of a
Beta-prior distribution with specific posterior risk to achieve minimum sample size, which avoids
the problem of estimating cost parameters. Hald (1981) has provided an excellent comparison on
classical and Bayesian theory and methodology for attributes Sampling Plans.
When Sampling plans are set for product characteristics that involves costly or destructive
testing by attributes, it is the usual practice to use Single Sampling plan with acceptance number
c=0 and c=1. But the OC curves for Single Sampling Plan with c=0 and c=1 leads to conflicting
interest between the producer and consumer. Here c=0 favours consumer and c=1 favours producer.
Govindaraju (1984) has proposed the Special Type of Double Sampling (STDS) Plan to avoid such
shortcomings
Model under consideration:
Let N and n denote the lot size and sample size X and x denote the number of defectives
in the lot and sample respectively.
The linear cost function is
I(N,n,X,x) = nS 1 + xS2 + (N-n) A1 + (X-x) A2 , for acceptance
= nS2 + xS2 + (N-n) R2 + (X-x) R2 , for rejection ..(1)
where nS 1 denotes the cost of inspection and xS 2 denotes the cost proportional to the number of
defective item in the sample. In fact S1 includes sampling and testing cost per item and S2 denotes
additional cost for an inspected defective item including the repair cost per item if in case the
defective items found in the sample which are repaired. Thus the cost n1 S1 + n2 S2 associated with
the sample provide costs for sampling inspection.
Cost of acceptance is given as (N-n) A1 + (X-x) A2 and Cost for outright rejection is (N-
n) R1 +(X-x) R2 .The parts (N-n) A1 and (N-n)R1 are proportional to the number of items in the
reminder of the lot and A1 and R2 are usually zero, generally negligible. The parts (X- x) A2
and (X-x)R2 are proportional to the number of defective items accepted and hence A2 and R1 are
often considerable.
From Hald (1981) the cost function stated in (1) becomes
I = n ( S1 + S2 p ) + (N-n) (A1 + A2 p), for acceptance ……….(2)
I = n (S1 +S2 p ) + ( N-n) (R1 +R2 p), for rejection ……….(3)
The average cost can be written as
K(N,n,n2 p)= n(S1 + S2 p) + (N-n)(A1 +A2 p) Pa(p)+[(R1 +R2 p )(1-Pa(p)] ….(4)
n n –1 ……
where Pa(p) = (1-p) + n2 p (1-p ) (5)
The double binomial distribution is a weighted average of two binomials with parameters p’ and
p’’,such that p’< p’’and weights w1 and w2 such that w1 +w2 =1.
Double binomial bayes solution
A necessary condition for a sampling plan to exist under the double binomial
distribution is that p’ < pr< p’’ where pr is a break-even quality. Assuming that p’ < pr<p’’, the
standardized cost function can be written as
R(N,n1 ,n2 ) = n+(N-n)[V1 (1-Pa(p’))+V2 Pa(p’’)] ……….(6)
Where V1 = w1 (pr – p’) and V2 = w2 (pr – p’’) ………. (7)
It is required to obtain (n1 ,n2 ) which minimizes equation (6). The necessary condition for such
sampling plan to exit is that V1 >0, V2 >0, p’<pr<p’’.
According to Hald (1981) the value for (n1 ,n2 ) minimizing R must satisfy the following
conditions
n 2 R(N,n,n2 -1) ≤ 0 ≤ n2 R ( N,n,n2 ) , 0<n2 < n ……..(8)
n R(N,n-1,n2 ) ≤0 ≤ n R ( N,n,n2 ) , 0<n < N ……..(9)
By substituting equation (5) in equation (6) and solving the inequality mentioned in equation (9)
one can get
n=-1/2+(ln(V2 /V1 )+ln(p’’/p’)+ln((1-p”+n2 p”)/(1-p’+n2 p'))+ln((1-p’)/(1-p’’)))/ln((1-p’)/(1-p”))
Similarly one can solve equation (8) for obtaining parameters. By giving various values for n2 , one
can compute the values for n with specified V1 ,V2 ,p’ and p’’.
Example: For specified p’=0.001, p’’=0.1,V1 =0.00891and V2 = 0.0009,and n2 = 15, the value of n
can be obtained as n=31 and n1 =[Link] the parameters of the STDS plan are n1 = 16 and n2 = 15.
REFERENCES
1. CALVIN,T.W.(1984): How and When to Perform Bayesian Acceptance Sampling, American
Society for Quality Control, Milwaunkee,Wisconsin,Vol.7
2. GOVINDARAJU,K. (1984):Contribution to the Studies of Certain Special Purpose Plan Ph.D.
Thesis , Bharathiar University, Coimbatore, TamilNadu , India.
3. HALD,A (1981): Statistical Theory of Sampling Inspection by Attributes, Academic Press
Inc.(London) Ltd.
4. OLIVER,L.R and SPRINGER M.D (1972). A General Set of Bayesian Attribute Sampling Plans,
American Institute of Industrial Engineers, Nov. Cross , GA.
RESUME
Ms. Deepa Menon is a research schilar mainly working in the field of acceptance sampling plans
and has presented papers in many international and national conferences in the field of product
control.
