# Estimated Standard Deviation

Six Sigma – iSixSigma › Forums › Old Forums › Europe › Estimated Standard Deviation

- This topic has 2 replies, 3 voices, and was last updated 12 years, 7 months ago by Scott.

- AuthorPosts
- November 27, 2007 at 2:47 am #23981
Hello,

If you know the Standard Deviation of the consumption of a part on a weekly basis, can you estimate the SD of the same part on a bi weekly basis ?

Example:

Weekly average consumption : 20 (calculated)

Weekly consumption Standard deviation : 5 (calculated)Bi weekly average consumption : 40 (calculated)

Bi weekly cosumption Standard deviation : ? (estimated)I have been told that the answer is 5/Sqr root of 2 but I am not convinced.

Please give me your thoughts

Thanks

Vincent0November 27, 2007 at 2:34 pm #57456

Adam L BowdenParticipant@Adam-L-Bowden**Include @Adam-L-Bowden in your post and this person will**

be notified via email.If your variation in demand (SD) does not change why would your

“number 5” change ?Often there is micro levels of seasonality hr to hr, day to day, week

to week, month to month etc which would provide for varying SD. But in your example if the week and bi-week variation in

consumption are the same the SD would not change.Any one else got some different opinions ?Regards,Adam0December 12, 2007 at 9:39 pm #57466Two Kudos:

1. For asking the question. So many people just assume it is half. Wrong.

2. For Adam’s response. On the right track

Can it be estimated, of course is yes, but you want some guidelines. Adam’s point about variables that can affect the dispersion bi-weekly that may not show on a weekly basis is right on. So, this is what makes the estimate more difficult. So many demand variations occur at month-end. This “Month-end” mentality suggests your question should be also which bi-weekly period within the month cycle.

Possibly in your favor is that you have the data to calculate the standard deviation correctly. If you have weekly data, then bin your data bi-weekly. 2. Rrun an SPC chart to look for stability in time series order. 3.Look for any pattern of “Month-end” cycles or other patterns that make the standard deviation vary bi-weekly to bi-weekly. 4. Also, one could use X Bar and R and subgroup on bi-weekly to see the patterns, if any, in the charts. 5. If stable and no patterns are evident, then calculate your standard deviations between these bi-weekly subgroups.

Now, I assume you thought of that, but you don’t really have the data. So, intuitively, you estimate the standard deviation over a longer period of time to be less compared to the mean demand than the standard deviation compared to the mean of a shorter period of time. 6. How much? Well, I would empirically answer this as you get data and make a stab at it now if you have no data at 70% of the coefficient of variation (St Dev/Mean) of the weekly. In other words, if the COV is 50% (Which I find often for high selling products) on a weekly basis, I estinate the standard deviation at 70% of the 50% COV for bi-weekly. But, I bet you have cycles, month-end and day of the week patterns (Especially on consumer luxury goods) that make this estimate tough. Let me know if you need more.0 - AuthorPosts

The forum ‘Europe’ is closed to new topics and replies.