# Sample Size Estimation

Six Sigma – iSixSigma › Forums › General Forums › Tools & Templates › Sample Size Estimation

- This topic has 15 replies, 5 voices, and was last updated 2 years, 4 months ago by Robert Butler.

- AuthorPosts
- October 27, 2013 at 11:50 pm #54571

Abhishek JainParticipant@abhi_jain80**Include @abhi_jain80 in your post and this person will**

be notified via email.Product validation group is facing a challenge in arriving at the right samples & right size of samples for test cases for certification. I need to build a sampling model which could determine the samples & samples size needed to release a product/patch to customers.

We have identified the factors that could impact the sample size. I am not getting the way how to start with. Which approach should I prefer.

Requesting the expert advice. Thanks in advance.0October 29, 2013 at 5:19 am #196134

Prabhu VParticipant@prabhuvspj**Include @prabhuvspj in your post and this person will**

be notified via email.Hello Mr. Jain,

Sampling technique needs what kind of data (continuous/discrete) you are going to handle at first.

Kindly elaborate your query.

Thanks and regards

Prabhu V.

0October 29, 2013 at 9:09 pm #196142

AbhishekGuest@abhi_jain80**Include @abhi_jain80 in your post and this person will**

be notified via email.Thanks Prabhu for the response.

The 2 factors identified are continuous, rest 4 factors are discrete and the response is binary i.e. Pass/Fail.

0October 30, 2013 at 10:22 pm #196148

Prabhu VParticipant@prabhuvspj**Include @prabhuvspj in your post and this person will**

be notified via email.Hello Mr. Jain,

Kindly find the below formulas for calculating sample size based on your inputs.

For continuous dataN= (Z*S/E)^2

N Sample size

Z – Constant for confidence level (like 1.645 90%, 1.96 95%, 2.575 99%)

S Standard deviation

E Precision or Error

(Precision or Error is basically difference between target and actual parameter like mean, yield etc)

Estimated Mean True Mean (like X bar µ)

Estimated Yield True YieldFor discrete data

N=(Z/E)^2*p(1-p)

E Precision or Error level of precision desired from the sample in units of proportion

p- ProportionPlease feel free to post your queries in case if you want for clarifications.

(Note: The above formulas are generic in nature, kindly support your sitaution for better clarity)

Thanks and regards

Prabhu V.

0November 4, 2013 at 4:07 am #196155Hi Prabu

Thanks for posting the formula for estimating sample. Can you plesae explain the below formula for discrete data with example. In my case we perform 89000 transaction per month. With this data how do we measure the sample size.

N=(Z/E)^2*p(1-p)

0November 4, 2013 at 11:29 pm #196165

Prabhu VParticipant@prabhuvspj**Include @prabhuvspj in your post and this person will**

be notified via email.Hello Baskar,

Kindly find the below guidelines based on your reply.

As you have not mentioned about your objective of sampling and its proportion, I am assuming your sampling objective is finding erroneous transactions in Population (89000 transactions) and erroneous proportion in Population is unknown currently (Assuming that you have 5% of transactions are erroneous).

Erroneous: 5% of transactions (p = 0.05)

Confidence level: 95% (Z = 1.96)

Sampling Precision or Error: 3% (E = 0.03)

N= (1.96/0.03)^2*(0.05(1-0.05)

N= (1.96/0.03)^2*(0.05*0.95)

N ~ 203 transactions

Please feel free to post your quires/clarifications if in needed on above.

Thanks and regards

Prabhu V.

0November 5, 2013 at 4:08 am #196166Thanks Prabu, for swift reply. The allowable error limit is 3% (SLA). Hence the p=0.03 right? then

N=(1.96/0.03)^2*(0.03*(1-0.03))

N=124

My basic question? 124 sampling is for day or month? if it is for month then why we do perform very low sampling when we process 89k transactions.Thanks for your support

0November 5, 2013 at 5:39 am #196167

MBBinWIParticipant@MBBinWI**Include @MBBinWI in your post and this person will**

be notified via email.@abhi_jain80 : I’m going to challenge your decision to have a pass/fail response variable. You are probably measuring something that is on a continuous scale and making a determination that it has passed or failed. If so, then you will be better off using that continuous scale rather than the discrete. This will allow you later to find the relationship of the inputs to the output which will then provide you the ability to improve the output value. Using a pass/fail gives you very limited ability to do this.

0November 8, 2013 at 3:00 am #196183

Prabhu VParticipant@prabhuvspj**Include @prabhuvspj in your post and this person will**

be notified via email.Hello Baskar,

On Addressing your query, I would like provide the following details

1) The above sample size guideline is not with respect to the Population, rather with respect to the Population proportion (This formula is for calculating the statistically significant discrete data sample size).

2) And also the above guideline did not guide you on the frequency or duration of samples from the Population.

3) While considering the sample frequency/duration you can consider on some other factors like current erroneous transactions, resource availability for collecting the samples, etc.

Please feel free to post your queries in case if you want any clarifications.

Thanks and regards

Prabhu V.

0November 13, 2013 at 5:59 am #196233Thanks Prabhu, it make sense.

Prabhu will you be able to assist me with different topic on “BENCH MARKING”. I have few clarification to verify with you

0November 13, 2013 at 9:10 pm #196236

Prabhu VParticipant@prabhuvspj**Include @prabhuvspj in your post and this person will**

be notified via email.Hello Baskar,

Thanks for your feedback.

You can very well post your query on any LSS (Lean & Six sigma) topics in this forum.

Definitely, it will be answered by someone who is having sufficient knowledge and expertise (of course if some suitable stuff with me means, surely will reply to your query)

If you want to contact me directly you can contact @ [email protected] or through LinkedIn

All the best!!!

Regards

Prabhu V.0November 14, 2013 at 10:52 pm #196242

Abhishek JainParticipant@abhi_jain80**Include @abhi_jain80 in your post and this person will**

be notified via email.Hi Prabhu,

I am from software industry & any software release has to go through test certification. Testing team basis on samples certify the release.

I need to develop the sampling model which could help them in identifying the right sampling size as well as right samples.

What should be my approach?? Appreciate your response.

Regards,

Abhi0November 15, 2013 at 3:40 am #196244

Prabhu VParticipant@prabhuvspj**Include @prabhuvspj in your post and this person will**

be notified via email.Hello Abhi,

As already mentioned that the sampling techniques will be decided based on the data type youre going to handle.

Hence if you can able to provide more details means, it will be useful to clarify your query.

For example as Mr. Baskar has explained about his situation.

Thanks and regards

Prabhu V.

0September 9, 2017 at 4:29 am #201820Please can you help me to use a formula to calculate a sample size from a population of 1,534 and explain to me vividly how you arrived at the answer. Thanks.

0September 9, 2017 at 7:53 am #201821

Katie BarryKeymaster@KatieBarry**Include @KatieBarry in your post and this person will**

be notified via email.Kofi – The more details you provide, the more likely you are to get a response. What do YOU think you should do? Why or why not? The iSixSigma audience is helpful, but they like to see that someone is putting forth a good-faith effort; they are not here to do your homework for you.

0September 9, 2017 at 10:03 am #201822

Robert ButlerParticipant@rbutler**Include @rbutler in your post and this person will**

be notified via email.As posed, your question has no answer other than to say that the minimum sample size could be 1 and the maximum sample size could be 1534. To provide anything meaningful to you it will be necessary for you to give us a detailed description of what it is that you are trying to do.

0 - AuthorPosts

You must be logged in to reply to this topic.