Sampling

Sampling using
Excel
and
Audit Commander

Audit Commander
Audit Guide
Data analysis made easier…
EZ-R Stats, LLC

Random Sequential Sampling

Audit Commander

The software described in this document may be freely downloaded
and used without restriction. Additional information about the audit
software is available at the web site http://ezrstats.com.
All comments and suggestions are welcome.


Document History

Revision History
Revision Number Revision Date Summary of Changes Author
1.0 8-8-2009 Initial Version M. Blakley
1.1 8-11-2009 Attribute sampling M. Blakley
1.2 8-21-2009 Stratified Sampling M. Blakley

Random Sampling User Guide Copyright 2009 Page 3 of 38


Table of Contents
1 ABOUT THIS GUIDE..........................................................................................................................6

1.1 W HO SHOULD USE IT........................................................................................................................6
1.2 TYPOGRAPHICAL CONVENTIONS ...........................................................................................................7
1.3 PURPOSE........................................................................................................................................7
1.4 SCOPE...........................................................................................................................................7
1.5 INTENDED AUDIENCE...........................................................................................................................8
1.6 HARDWARE REQUIREMENTS..................................................................................................................8
1.7 SOFTWARE REQUIREMENTS..................................................................................................................8

2 GETTING STARTED..........................................................................................................................9

2.1 RANDOM SEQUENTIAL SAMPLING............................................................................................................9
2.2 ORGANIZATION OF THE GUIDE.............................................................................................................10
2.3 SAMPLING OBJECTIVES......................................................................................................................10

3 SEQUENTIAL SAMPLING WITH EXCEL (UNRESTRICTED).........................................................11

3.1 DESCRIPTION OF DATA TO BE SAMPLED..................................................................................................11
3.2 VARIABLE SAMPLING.........................................................................................................................13
3.3 ATTRIBUTE SAMPLING.......................................................................................................................18

4 SEQUENTIAL SAMPLING WITH AUDIT COMMANDER (UNRESTRICTED).................................20

4.1 MACRO USED FOR VARIABLE SAMPLING...............................................................................................20
4.2 TEXT REPORT................................................................................................................................21
4.3 CHART PRODUCED..........................................................................................................................22
4.4 MACRO USED FOR ATTRIBUTE SAMPLING..............................................................................................23
4.5 TEXT REPORT................................................................................................................................23
4.6 CHART PRODUCED..........................................................................................................................25
4.7 COMPUTING UPPER AND LOWER LIMITS USING A MACRO...........................................................................26

5 STRATIFIED SAMPLING WITH EXCEL .........................................................................................28

5.1 VARIABLE SAMPLING........................................................................................................................28
5.2 STRATIFIED ATTRIBUTE SAMPLING........................................................................................................32

6 STRATIFIED SAMPLING WITH AUDIT COMMANDER ..................................................................34

6.1 MACRO USED FOR VARIABLE SAMPLING...............................................................................................34
6.2 TEXT REPORT................................................................................................................................35

MACRO USED FOR STRATIFIED ATTRIBUTE SAMPLING...................................................................................36
6.3 OUTPUT REPORT............................................................................................................................37



1 About this guide

This document is divided into the following chapters:

• Chapter 1 – Overview

• Chapter 2 – About sequential sampling

• Chapter 3 – Using Excel For Random Sequential Sampling (Unrestricted)

• Chapter 4 – Random Sequential Sampling with Audit Commander (Unrestricted)

• Chapter 5 –Stratified Sampling using Excel

• Chapter 6 –Stratified Sampling with Audit Commander

1.1 Who Should Use It
Auditors, researchers, business analysts and academics who use random sampling to perform
their jobs.

• Auditors: can use the software to sample more efficiently

• Researchers: use the software for:

• Random sampling

• Statistical reports and charts


1.2 Typographical Conventions
This document uses the following typographical conventions:

• Command and option names appear in bold type in definitions and examples.

• Screen output and code samples appear in monospace type.

1.3 Purpose
The purpose of this monograph is to provide a practical guide to the use of random sequential
sampling in order to make the sampling process more efficient.

Random sequential sampling is not a new concept, but has received very little attention. This
may be due to the fact that to take maximum advantage of sequential sampling, an automated
process is necessary. Hopefully, this guide will provide the information to do so.

The auditor does not need special computer skills in order to be able to take advantage of
random sequential sampling.

Development of the software began in August 2005 when the author searched fruitlessly for a
relatively easy to use, economical software package for random sampling. During its
development, suggestions and improvements were made by a variety of audit practitioners.

More information about the system is available from the website, http://ezrstats.com.

1.4 Scope
The purpose of this guide is to explain the basis for sequential sampling and methods to
implement it using software. This guide is for version 1.88 of the Audit Commander software.



1.5 Intended audience
The software is intended for use by both internal and external auditors, researchers, students
learning random sampling, business analysts and anyone else interested in performing random
sampling in a more efficient and effective manner.

1.6 Hardware requirements
At least 512 MB of memory (more if possible). Minimum disk space is 27 MB.

1.7 Software requirements
Works only in Windows XP, Vista or Windows 7. Requires ActiveX Data Objects which is part of
SP1. (ActiveX Data Objects can be downloaded from the Microsoft web site at no charge)


2 Getting Started

2.1 Random sequential sampling
Sequential sampling differs from the more traditional methods of performing random sampling,
yet provides the same results. In sequential sampling, there does not need to be any “up front”
analysis or knowledge of the attributes of the data being sampled. There is no calculation of
“required sample size”. Instead, sampling objectives are defined at the start of the process, and
then small, “sequential” samples are taken and evaluated. If, after evaluating the cumulative
results, the sampling objective has been achieved, the sampling process ends. If the sampling
objective has not been achieved, then another small sample is taken and the process is
repeated.
The advantages of sequential sampling include:

• No knowledge of the attributes of the population being tested are required

• No need to compute a “required sample size”

• No need to pre-select “spares” in the event that the sample size taken is too small

• Little or no chance of “over sampling” and thus spending unneeded time testing the
samples

The disadvantages of sequential sampling include:

• Not applicable or practical in all instances

• Computationally intensive

However, in many cases the advantages of sequential sampling outweigh the disadvantages.



2.2 Organization of the guide
A small sample test dataset has been prepared and will be used for illustration of the sampling
procedures. Two types of sampling will be illustrated:
1. Attribute sampling
2. Variable sampling

The sampling procedures will be illustrated using two software systems:
1. Microsoft Excel
2. Audit Commander 1.88

2.3 Sampling objectives
The sampling objectives will be identical for all three systems used. For the variable sample, a
confidence level of 95% is desired, with a precision level of 5%.
For the attribute sample, a confidence level of 95% is also desired, with the same precision level
of 5%. Note that although the examples use these confidence levels, any confidence level may
be used.


3 Sequential Sampling with Excel
(Unrestricted)

3.1 Description of data to be sampled
For this exercise, example data has been included on two Excel worksheets. One sheet is for
variable sampling and is named “Sample Variable”. The other sheet is for attribute sampling and
has been named “Sample Attribute” .
Shown below is a snapshot of the top of each sheet.

3.1.1 Variable sample data on worksheet “Sample Variable”

Book Audited
Obs No Sampled Value Value Random Tran Date Actual Value
1 209.37 0.1404 7/10/07 209.37
2 512.48 0.0341 5/23/07 512.48
3 324.26 0.1126 6/16/07 324.26
4 749.61 0.0428 5/19/07 749.61
5 103.9 0.6257 5/19/07 103.9
6 635.67 0.3944 5/3/07 635.67
7 123.25 0.2944 6/25/07 123.25
8 665.44 0.9755 7/5/07 620.91
9 540.02 0.5047 4/13/07 540.02
10 727.32 0.56 6/7/07 727.32
11 237.54 0.2611 6/21/07 237.54
12 104.63 0.7042 4/18/07 108.37
13 796.32 0.2072 7/7/07 796.32
14 864.74 0.3659 6/27/07 864.74
15 811.11 0.1485 4/16/07 811.11
16 142.15 0.9427 6/21/07 145.12
17 295.03 0.0622 7/10/07 295.03

As can be seen, the worksheet consists of seven columns as follows:

• Obs no – the observation number or assigned number for the population being
tested.


• Sampled – a blank column indicating whether the row has been selected for
sampling

• Book Value – the recorded numeric amount to be tested

• Audited value – the amount that the transaction should have been, based upon the
auditor’s audit tests performed when the item was sampled and reviewed

• Random number – a random number between 0 and 1 that was assigned using the
Excel random function “=rand()”

• Tran date – a transaction date, for reference purposes only

• Actual value – this is the value that would be obtained during an audit, i.e. the
audited value. Note that in an actual audit this information would not be available,
but is shown here only for educational and illustrative purposes.

3.1.2 Attribute sample data on worksheet “Sample Attribute”

Obs No Sampled Signature Document Approval Random Tran Date
1 0.05376 7/10/07
2 0.0341 5/23/07
3 0.1126 6/16/07
4 0.0428 5/19/07
5 0.6257 5/19/07
6 0.3944 5/3/07
7 0.2944 6/25/07
8 0.9755 7/5/07
9 0.5047 4/13/07
10 0.56 6/7/07
11 0.2611 6/21/07
12 0.7042 4/18/07
13 0.2072 7/7/07
14 0.3659 6/27/07
15 0.1485 4/16/07
16 0.9427 6/21/07
17 0.0622 7/10/07
18 0.863 5/27/07

As can be seen, the worksheet consists of seven columns as follows:

• Obs no – the observation number or assigned number for the population being
tested.

• Sampled – a blank column indicating whether the row has been selected for
sampling

• Signature – A blank column used to record if the selected attribute “signature” exists
and was appropriate

• Document – A blank column used to record if the selected attribute “signature” exists
and was appropriate

• Approval – A blank column used to record if the selected attribute “signature” exists
and was appropriate

• Random – A random number between 0 and 1 generated using the Excel function
“=rand()”

• Tran date – A date used for reference purposes only.

3.2 Variable sampling

3.2.1 Sampling the first group

The first step in the procedure is to define the sampling objectives, which are the same for all
sampling systems tested, i.e. 95% confidence and 5% precision. The next step is to assign a
random number between 0 and 1 to each item in the population. In this example, the Excel
random number function has been used, although it is possible to use other methods to assign
random numbers.
The next step is to sort the data by the random numbers assigned. Order may be either
ascending or descending. For purposes of this illustration we select ascending order.
The next step is to select a small starting number of transactions. The selections must be made
sequentially based upon the sorted items to be tested in the population.
We will select the first 150 items and then make an audit determination of what the audited value
is, i.e. after reviewing the documents and performing other tests on the items in the sample
make a determination of the audited values for each of the first 150 items.


Note that for this example, the “audited value” has already been specified in the rightmost
column. So we can copy those values over to the column “Audited Value”. The worksheet now
appears as follows:

Book Audited
Obs No Sampled Value Value Random Tran Date Actual Value
232 X 979.08 979.08 0.001 6/22/07 979.08
432 X 838.61 838 0.002 5/28/07 838.61
253 X 64.65 64.65 0.0044 6/10/07 64.65
368 X 272.39 270 0.0053 4/30/07 272.39
405 X 334.11 0 0.0085 7/11/07 334.11
221 X 115.73 110 0.0108 5/16/07 115.73
270 X 355.41 0 0.0108 7/16/07 355.41
227 X 32.37 32.37 0.0161 4/22/07 32.37
288 X 140.56 140 0.0191 4/23/07 140.56
34 X 987.76 987.76 0.0252 4/23/07 987.76
132 X 285.93 270 0.0256 6/28/07 285.93
360 X 169.05 169.05 0.0259 6/13/07 169.05
313 X 569.69 560 0.0268 7/10/07 569.69
298 X 855.35 855.35 0.0269 6/9/07 855.35
345 684.14 0.0322 6/22/07 684.14
216 901.32 0.0328 5/27/07 901.32
2 512.48 0.0341 5/23/07 512.48

Our audited values are now contained in column “D”, rows 2 through 151.

3.2.2 Assessing the first group

The first group of sampled transactions has their audited values in column “D”.
We can use the Excel function “STDEV” to determine the sample standard deviation. To simplify
and clarify the formulae used, set up a “named range” for the sampled values. In this instance,
the sampled values range will be called “sampled” and is defined as the cells from D2:d151.
Similarly, a named range “population” is established for the entire population. The formula for the
standard deviation is then “=STDEV(sampled)”. We an use the COUNT function to determine
the number of items in the population. The formula is “=COUNT(population)”. We can also
determine the average (mean) value for the sampled items using the Excel AVERAGE function –
for which the formula is “=AVERAGE(sampled)”.

The point estimate of the value for the entire population will simply be the average times the
number of items which can be obtained by multiplying the two values we obtained, i.e. use the
formula “=AVERAGE(sampled) * COUNT(population)”.

The confidence interval at 95% confidence is the Student’s t-value using the sample count – 1
degrees of freedom. In Excel, the formula for the Student’s t-value would be
“=TINV(.05,count(sampled)-1)”. The standard error of the mean is the sample standard
deviation divided by the square root of the sample size or
“STDEV(sampled)/SQRT(COUNT(sampled))”. A 95% confidence interval would then be
obtained by multiplying this amount by the population count (population) and multiplying that by
the t-value (TINV(.05,count(sampled)-1). For samples which represent more than 5% of the
population size, a finite population correction factor must be applied. This factor is the square
root of the population size minus the sample size divided by the population size minus one. In
Excel, the formula is “=SQRT((COUNT(population) – COUNT(sampled))/(COUNT(population)
– 1))”. So the formula for the interval would be:
“=(STDEV(sampled)/SQRT(COUNT(sampled)))* TINV(.05,count(sampled)-1) *
count(population)* SQRT((COUNT(population) – COUNT(sampled))/(COUNT(population) –
1”.

Once all the formulae are established, it becomes very easy to compute all these amounts. All
that is required is changing the specification for the named range “sampled”.

Our first test is for 150 items. Once the samples are evaluated (audited), the following results
are obtained:

Mean 494.36
Std Dev 283.23
Count 150
Pop 432.00
PE 213,563
Corr 0.81
SE Mean 23.13
t-value 1.97601
Interval 15,968.37
Percent 7.48%
LL 197,595
UL 229,531



The confidence interval percentage is 7.48% which is too wide. Thus we need to test more
samples. Shown below are the results for the selection of an additional 85 samples (235 in
total). Note that it is fairly easy to do the calculations in Excel, as the only thing that needs to be
changed is the definition of the sampled range, i.e. change from D2:d151 to d2:d236.

The results achieved are as follows:

Mean 491.94
Std Dev 283.62
Count 235
Pop 432.00
PE 212,519
Corr 0.68
SE Mean 18.50
t-value 1.97015
Interval 10,645.98
Percent 5.01%
LL 201,873
UL 223,165

At this point we achieved our sampling objectives, i.e. a precision level of 5% has been obtained.



3.3 Attribute Sampling

Evaluating the results of attribute sampling using Excel is difficult, to say the least. An estimate
for the results can be computed using the following formulae. The error rate is computed as the
number of errors found in the sample divided by the sample size, i.e. (errors_found /
sample_size). The point estimate is simply the error rate (errors_found / sample_size) *
population_size. The error rate is generally notated as “p”.

The sample size is generally designated as “n”. The population size is generally designated as
“N”. The formula for the universe standard error is “=SQRT((p*(1-p))/(n-1) * (1 – (n/N))”.
A “z-score” can be computed using Excel’s inverse normal function. For a 95% precision, the
formula would be “=NORMINV(.975,0,1)” which results in a value of 1.959963.

To compute the upper limit, determine the interval size as the universe standard error times the
z-score added to the point estimate. For the lower limit, simply subtract the interval size from the
point estimate.

The formula is shown below:

However, this technique yields only approximations, which may have significant errors in certain
cases. The exact approach, which is used by RAT-STATS, is based on the formulae published
in the following article.John P. Buonaccorsi (1987), “A Note on Confidence Intervals for
Proportions in Finite Populations,” The American Statistician, Vol. 41, No. 3, 215-218.

The formula is computationally intensive and requires the use of the mathematical
“Combinations” function. Although Excel has the function (“=COMBIN(m,n)”), the process to
make the computations would be difficult to perform using a spreadsheet alone.


Bottom line – Excel can compute only a rough estimate of attribute sampling values.

The formulae referenced in the article to compute the upper limit is:

The computation for the lower limit is similar, as shown below:



4 Sequential Sampling with Audit Commander
(Unrestricted)

Sequential sampling with Audit Commander involves opening a form and selecting values from
the drop down menus. Note that no “pre sorting” is required, and it Is possible to specify both
the work sheet and the columns to be tested.

4.1 Macro Used for Variable Sampling

Dim oewm As New cEWM

Dim sInfile As String

Dim sSQL As String

sInfile = "c:testsampleExample data.xls"

oewm.ChartFile = "c:testsamplet-stopgov.png"

oewm.Cmd = "stopgov"

oewm.Colstring = "[sampled], [audited value]"

oewm.Table = "Sample Variable$" ' req by oewm

'oewm.SQL = "select county, mileage, expense from [maintenance$];"

oewm.Filein = sInfile

oewm.FileOut = "c:testsamplet-stopgov-report.txt"

'oewm.Parm1 = ".95"

oewm.Title = "Test Stop Go V"

'MsgBox oewm.ShowValues

Call oewm.RunCmd

If oewm.ErrNo <> 0 Then

MsgBox oewm.errmsg

MsgBox oewm.ErrDescription

End If

Set oewm = Nothing

4.2 Text Report
The following text report is output by Audit Commander to summarize the results of the
sampling.

Sampling results:
Sample size 64
Sample mean 460.05
Sample Std Dev 302.48
Population size 432
Point estimate: 198,740.99

Values at 95% confidence
t-value used 1.99834
Lower limit 168,579.74
Upper Limit 228,902.24
t-value 1.99834

t-value 2.38701

t-value 1.66940

t-value 1.29513



4.3 Chart Produced
The following chart is output by the Audit Commander to show the results of sampling as a chart.
These results reflect the intervals at four levels of confidence – 80%, 90%, 95% and 98%
confidence.

The chart shows the ranges of values. On the left, the upper and lower intervals at the top and
bottom are for the 98% confidence levels. The two lines between are the upper and lower
intervals at the 95% confidence level.
The chart on the right side of the report shows the values at 80% and 90% confidence levels.
The outside upper and lower confidence intervals are the values at the 90% confidence level,
with the values inside showing the amounts at the 80% confidence level.


4.4 Macro Used for Attribute Sampling
Audit Commander uses the same formulae as RAT-STATS for attribute sampling, i.e. the formula
published in the Journal article previously mentioned.


Dim sInfile As String

Dim sSQL As String

sInfile = "c:testsampleExample data.xls"

oewm.ChartFile = "c:testsamplet-stopgoa.png"

oewm.Cmd = "stopgoa"

oewm.Colstring = "[sampled], [Signature]"

oewm.Table = "Sample Attribute$" ' req by oewm

oewm.Filein = sInfile

oewm.FileOut = "c:testsamplet-stopgoa-report.txt"

oewm.Title = "Attribute testing of signatures"

Call oewm.RunCmd

If oewm.ErrNo <> 0 Then

MsgBox oewm.errmsg

MsgBox oewm.ErrDescription

End If

Set oewm = Nothing

4.5 Text Report
The following text report is output by Audit Commander to summarize the results of the attribute
sampling.



Sampling results:
Sample size 98
Errors 4
Error rate 4.08%
Population size 3027
Confidence used 95.00%
Z-score 1.95996
Point estimate: 124
Lower limit 35
Upper Limit 303
Lower limit 27
Upper Limit 341
Lower limit 44
Upper Limit 272
Lower limit 55
Upper Limit 239


4.6 Chart Produced
The following chart is output by the Audit Commander to show the results of sampling as a chart.
These results reflect the intervals at four levels of confidence – 80%, 90%, 95% and 98%
confidence.

The chart shows the ranges of values. On the left, the upper and lower intervals at the top and
bottom are for the 98% confidence levels. The two lines between are the upper and lower
intervals at the 95% confidence level.
The chart on the right side of the report shows the values at 80% and 90% confidence levels.
The outside upper and lower confidence intervals are the values at the 90% confidence level,
with the values inside showing the amounts at the 80% confidence level.



4.7 Computing Upper and Lower Limits using a
macro
If only the limits need to be computed (i.e. no report or charts), the upper and lower limits can be
computed directly using a macro such as shown below:

Dim nsamp As Long

Dim nPop As Long

Dim nErrs As Long

Dim dPrec As Double

Dim iuL As Long

Dim iLL As Long

Dim sMsg As String

Dim oEWM As New cEWM

nsamp = 98

nPop = 3027

nErrs = 4

dPrec = 0.8

iuL = oEWM.ul(nsamp, nPop, nErrs, dPrec)

iLL = oEWM.ll(nsamp, nPop, nErrs, dPrec)

sMsg = "Attribute testing: " & vbCrLf

sMsg = sMsg & "Population: " & nPop & vbCrLf

sMsg = sMsg & "Sample size: " & nsamp & vbCrLf

sMsg = sMsg & "Errors: " & nErrs & vbCrLf

sMsg = sMsg & "Precision: " & dPrec & vbCrLf

sMsg = sMsg & "LL: " & iLL & vbCrLf

sMsg = sMsg & "UL: " & iuL & vbCrLf

Set oEWM = Nothing

MsgBox sMsg



5 Stratified Sampling with Excel

5.1 Variable Sampling
The process to perform stratified variable assessments of sample results in Excel is similar to
that for unrestricted sampling. The primary difference is that additional computations need to be
made for each strata.

5.1.1 Point estimate

The end objective is the same, i.e. determine a point estimate and determine a confidence
interval. The point estimate for stratified variable sampling is simply the sum of the point
estimates for each strata.

5.1.2 Population Variance

The estimated variance for the population is computed using the following formula:

Although this formula looks imposing, it can be broken down into small components, each of
which is readily calculated using Excel. The starting point is the calculation of the sample
variance for each stratum. To compute this amount using Excel, a series of columns can be set

up with each strata shown on a separate row. Begin by computing the count, average and
standard deviation for each strata using the following built in Excel functions:
Statistic Excel function Symbol in formula
Count “=Count(range)” nI
Average “=Average(range)” Y bar i
Standard Deviation “=Stdev(range)” si

Now that we have the standard deviation and counts for each strata, we can then perform the
computations step by step. In the Excel sheet shown below, the first column is the stratum
name or number, the second column is the number of items sampled from each stratum followed
by the average and standard deviation. Next is the count of all items in the stratum.

The remaining columns are completed as follows.
The point estimate (PE) is simply the average times the number of items in the entire stratum.
The remaining columns are used to compute the value of the formula which can be expressed
as estimated population variance = A * sum( B * C * D) where “A” is 1 / N squared, “B” is the
number of items in each strata etc.
Thus the three columns to the right of the column labeled “PE” are the “B”, “C” and “D” of the
formula. Once we have each of these three amounts, we can then use Excel to multiply them
together. This provides the value in the right most column. All of the values in the rightmost
column can then be totaled and this total divided by the entire population count (N) squared.

Strata n Average Std dev N PE s1^2*/ni (Ni-ni)/Ni) Ni^2 G*H*I
1 49 8590.51 17056.93 51 438115.78 5937527.39 0.04 2601 605627793
2 30 1486.85 2541.63 267 396989.82 215330.23 0.89 71289 13625881685
3 124 563.00 870.69 4,512 2540253.45 6113.77 0.97 20358144 121044444083
4 30 210.14 267.69 1,724 362279.06 2388.67 0.98 2972176 6976000315
5 30 179.64 187.17 2,074 372578.20 1167.81 0.99 4301476 4950657288
6 30 19.15 42.21 4,825 92416.12 59.38 0.99 23280625 1373775814
7 30 11.43 23.55 7,114 81336.73 18.48 1.00 50608996 931341609
2 12 1674.00 1283.20 12 20087.97 137215.90 0.00 144 0
3 30 226.67 501.19 1,489 337513.36 8372.93 0.98 2217121 18189788402
4 30 174.20 273.09 1,922 334805.35 2486.01 0.98 3694084 9040185400
5 30 54.46 120.32 5,539 301672.40 482.54 0.99 30680521 14724396938
6 31 33.14 51.05 24,790 821430.24 84.07 1.00 614544100 51597546090
7 47 21.47 27.19 72,369 1554104.26 15.73 1.00 5237272161 82323327630

Totals 503 126,688 7653582.75 325382973046
20.27327273
60.4128469 Square root 4.502585116
Z -1.959962787
Interval -8.824899275

LL 51.58794765 6535574
UL 69.2377462 8771592

In this example this result is 20.2732. This is the population variance.



To obtain our confidence limits, we would then use the following formula:

The square root of the population variance is 4.502585. The z-score can be computed using the
Excel formula “=NORMINV((1+conf)/2,0,1). For a confidence of 95% this yields a value of
1.95996. Thus, in this example, the interval would be 8.8248 which is the product of the two
amounts. This amount would then be added to (or subtracted from) the average value. These
amounts would then be multiplied by the entire population counts (in this example 126,698 to
arrive at the upper and lower confidence levels (here 8,771,592 and 6,535,574.
In the example shown, the sample sizes were selected arbitrarily. In some cases it may be
preferable to optimize these sample sizes in order to obtain the greatest precision for the least
sample size.

5.1.3 Optimizing the sample

In some cases it is desirable to optimize the number of samples selected from each strata in
order to get the most “bang for the buck” if the number of samples to be taken is a fixed, pre-
determined number.
There are at least two types of optimization that can be performed:
1. The cost for the selection of a sample in a strata is known, or
2. The cost is assumed to be the same for each strata (or unknown)

The formula to compute the optimal number of samples for each strata when the cost is known
is as follows:

The computations can readily be done using Excel by setting up multiple columns.

Stratum N Std Dev Sample Cost Nh * sh Sqrt(ch) E/F Optimal
Wake County 5367 21.65 1 116195.55 1.00 116195.55 32
New Hanover 345 18.65 3 6434.25 1.73 3714.82 1
Mecklenberg 756 19.21 2 14522.76 1.41 10269.14 3
Person 24 24.56 1.5 589.44 1.22 481.28 0
Orange 654 17.23 1.5 11268.42 1.22 9200.63 3
Durham 1298 45.32 2 58825.36 1.41 41595.81 11

Totals 11 181457.22 50

Required Sample size 50

If the cost is the same for all strata then the following formula can be used:

An alternative is to use the same formula and then just use a cost of ‘1” for all strata.

Stratum N Std Dev Sample Cost Nh * sh Sqrt(ch) E/F Optimal
Wake County 5367 21.65 1 116195.55 1.00 116195.55 28
New Hanover 345 18.65 1 6434.25 1.00 6434.25 2
Mecklenberg 756 19.21 1 14522.76 1.00 14522.76 3
Person 24 24.56 1 589.44 1.00 589.44 0
Orange 654 17.23 1 11268.42 1.00 11268.42 3
Durham 1298 45.32 1 58825.36 1.00 58825.36 14

Totals 6 207835.78 50

Required Sample size 50



5.2 Stratified Attribute Sampling

5.2.1 Point estimate

The end objective is the same, i.e. determine a point estimate and determine a confidence
interval. The point estimate for stratified variable sampling is simply the sum of the point
estimates for each strata. The point estimate for a strata is the percentage of attributes
identified in the sample multiplied by the number of rows in the strata, and rounded to the
nearest integer. The point estimate for the entire population is then the sum of all the point
estimates for each strata.

A determination of the point estimate using Excel is illustrated in the figure below:

Point
Stratum N n Count p Estimate
Wake County 5,367 57 4 7.02% 377
New Hanover 345 6 1 16.67% 58
Mecklenberg 756 15 0 0.00% 0
Person 24 2 1 50.00% 12
Orange 654 3 0 0.00% 0
Durham 1,298 15 2 13.33% 173

Totals 8,444 98 8 620

Point estimate 7.34%

The formula used is:

5.2.2 Population Variance

The estimated variance for the population is computed based upon the estimated sample
variance for each strata using the following formula:

The overall population variance is then estimated using the following formula:

These amounts can be computed readily in Excel. The process is best illustrated by breaking up
the formula into small components and then combining the amounts to obtain the final
calculation.

An illustration of the process to compute the population variance using Excel is illustrated in the
figure below.

Fraction Square root
Stratum N n Count p (1-p) (E*F)/(C-1) (N-n/N) (G*H) (N/Total) Product
Wake County 5,367 57 4 7.02% 92.98% 0.12% 98.94% 3.40% 63.56% 2.16%
New Hanover 345 6 1 16.67% 83.33% 2.78% 98.26% 16.52% 4.09% 0.68%
Mecklenberg 756 15 0 0.00% 100.00% 0.00% 98.02% 0.00% 8.95% 0.00%
Person 24 2 1 50.00% 50.00% 25.00% 91.67% 47.87% 0.28% 0.14%
Orange 654 3 0 0.00% 100.00% 0.00% 99.54% 0.00% 7.75% 0.00%
Durham 1,298 15 2 13.33% 86.67% 0.83% 98.84% 9.03% 15.37% 1.39%

Totals 8,444 98 8 4.36%

Standard Error 4.36%
Confidence 95.00%
Z-score 1.95996
Interval 0.08541
Point estimate 8.16%
Upper 16.70%
Lower -0.38%

When the lower limit is computed as less than zero, it is then changed to zero.



6 Stratified Sampling with Audit Commander

Sequential sampling with Audit Commander involves opening a form and selecting values from
the drop down menus. Note that no “pre sorting” is required, and it Is possible to specify both
the work sheet and the columns to be tested.

6.1 Macro Used for Variable Sampling


Dim sMsg As String

Dim sColstring As String

oewm.Filein = "c:testewpNeyman v3.xls"

oewm.Cmd = "svacc"

oewm.FileOut = "c:testcmreportt-33.txt"

sColstring = "[examined]"

sColstring = sColstring & ",[stratum]"

sColstring = sColstring & ",[selected]"

oewm.Colstring = sColstring

oewm.LocalCol = "stratum"

oewm.Table = "Neyman$"

'number of columns

oewm.Parm1 = "45"

oewm.Parm2 = ".95"

oewm.RunCmd

6.2 Text Report
The following text report is output by Audit Commander to summarize the results of the
sampling. (Txt report has been imported into Excel for ease of viewing).

Standard
Strata N n Mean Deviation Point Estimate Lower Limit Upper Limit
1 345 2 62.46 0 21,548.70 21,548.70 21,548.70
2 337 2 113.4 0 38,214.12 38,214.12 38,214.12
3 696 7 295.5 55.04 205,670.98 130,586.75 280,755.21
4 1431 15 511.84 29.35 732,443.04 650,134.56 814,751.52
5 2213 32 629.45 70.57 1,392,962.48 1,086,882.71 1,699,042.24
6 691 13 895.59 62.84 618,853.22 703,954.90 533,751.55
All 5713 71 589.9 54.94 3,370,088.24 2,754,943.74 3,985,232.74

Neyman Allocation report
Strata N Std Amt Pct Samp Size Next
1 345 28.64 9,880.94 3.36% 2 -343
2 337 29.2 9,841.66 3.35% 2 -335
3 696 59.11 41,138.89 13.99% 6 -690
4 1431 36.96 52,886.43 17.99% 8 -1,423
5 2213 63.48 140,484.00 47.78% 21 -2,192
6 691 57.62 39,817.29 13.54% 6 -685



Macro Used for Stratified Attribute Sampling
Audit Commander can assess stratified attribute sampling using either a menu or a macro. The
macro used is shown below (which provides the same results as the form based system).


Dim sMsg As String

Dim sColstring As String

oewm.Filein = "c:testewpNeyman v3.xls"

oewm.Cmd = "saacc"

oewm.FileOut = "c:testcmreportt-34.txt"

sColstring = "[audited]"

sColstring = sColstring & ",[stratum]"

sColstring = sColstring & ",[selected]"

oewm.Colstring = sColstring

oewm.LocalCol = "stratum"

oewm.Table = "Attrib$"

oewm.Parm2 = ".95"

oewm.RunCmd

If oewm.ErrNo = 0 Then

Application.StatusBar = "OK: " & oewm.FileOut

'MsgBox oEWM.Sql

Else

sMsg = "Errno: " & oewm.ErrNo & vbCrLf

sMsg = sMsg & "Errtext: " & oewm.ErrDescription & vbCrLf

MsgBox sMsg

End If

Set oewm = Nothing

6.3 Output Report
The following text report is output by Audit Commander to summarize the results of the attribute
sampling. (text report has been imported into Excel for ease of viewing).

Standard Point Lower Upper
Strata N n Mean Deviation Estimate Limit Limit
1 594 17 17.65% 0.91% 105 94 115
2 583 17 5.88% 0.35% 34 30 38
3 1132 12 8.33% 0.69% 94 79 110
4 863 12 0.00% 0.00% 0 0 0
5 1399 12 0.00% 0.00% 0 0 0
6 1142 12 0.00% 0.00% 0 0 0
All 5713 82 2.86% 0.27% 164 134 194



Comment form

Please send any comments, suggestions or items identified as errors to:

Mike.Blakley@ezrstats.com

Although I am not able to respond to all such comments and suggestions, I will try
to do so as feasible. Registered users of Audit Commander will be notified as
revised versions of the manual are released.

Sampling

More Related Content

Viewers also liked (8)

Similar to Sampling (20)

Recently uploaded (20)

Sampling