When are Statistical techniques not applicable?

[Marc]

Re: When are Statistical techniques not appplicable?

I appreciate your clarifying it for me, Bev and Steve. When I think of SPC I think of high volume processes (e.g.: "your time series data") rather than more broadly. I guess one could use SPC if one data point is made every week and data is analyzed over 5 or 10 years, or am I way off base? Obviously I'm not a statistician nor am I a mathematician.

I *think* you did support my initial short reply in this discussion thread - SPC is just one of many "statistical techniques". And I totally agree that "gut feel" isn't appropriate in any situation.

 

[Ajit Basrur]

Re: When are Statistical techniques not appplicable?

ISO 10007:2003 Guidance on statistical techniques for ISO 9001:2000 offers 12 high level types of statistical techniques. Check them at the left column in the standard preview page.

Small correction - should be ISO /TR 10017 and not 10007

 

[Steve Prevette]

Re: When are Statistical techniques not appplicable?

Real life story - back at Hanford, the radiological controls group at a facility was making about a dozen charts from a log for their management. It was about 40 hours a month (one week) that this person was putting into make the monthly charts. He was counting the data up by hand, making pie charts, and then management would return with - can you make them 3-D? Can you make the background mauve? Meanwhile the charts said nothing useful and no actions were taken.

Out of frustration they gave the process to me in the central group. I was able to take the log (which was in Excel) make the data calculations automatically, and set up the SPC charts. It took 15 minutes from the time that I received the email with the log to when I returned the charts, and no, requests for Mauve backgrounds were denied.

I use this story when people ask me - how expensive is it to shift to SPC.

By the way, Japanese high school students learn how to make SPC charts. There was a secretary (no college degree) that volunteered at Hanford to make charts, she made 300 charts per month with minimal supervision.

 

[Marc]

Re: When are Statistical techniques not appplicable?

That is a big part of the problem in any organization, which is it can be hard to find someone who knows the theory and can use the tools. I can see why Matlab, for example, is of little value to anyone but a math person. As to Excel - Just like Word - It is quite powerful. The problem is to find someone who knows Excel (not to mention statistics) in depth, including programming in Excel.

I first used Excel in the late 1980's when it came out for the Mac. There wasn't even a Windows version of Excel out. But - While I can do some basic stuff, even after all these years I can say I do not *know* Excel. I use it for a lot of things, even now that I'm retired, but I was never in a position where I had to use it for much other than simple stuff. I'd say it shows I'm not curious, or something, but since the 1980's I have not been in a position where I had a need to learn how to program in Excel, make macros, etc., or deal a lot with statistics beyond the "normal" stuff.

The problem becomes acute in smaller companies where they do not have any personnel who are good with statistics. Many of them don't have anyone who is even good with Excel, which I consider a basic, even for data entry. I have a friend who ran a group of restaurants. They're gone now, but I remember not so many years ago he had a hard time finding managers for each "store" who could reliably input daily data into a pre-made Excel spreadsheet. I will say, in retrospect, that is, in my opinion, part of why his business venture eventually failed. He didn't keep his eyes on his data in many areas. And I told him so way back around 2005 before things started going down hill (2008ish things started to go "bad"). I had invested in his business by way of some loans. I got him to pay me off and got out or it (essentially I was an investor) back in 2005-6. I wouldn't even consider someone who isn't at least minimally "Excel knowledgeable".

I have a lot of respect for people who are good in math. It never was my strong point.

 

[John Broomfield]

Re: When are Statistical techniques not appplicable?

I just came from a Lead Assessor Training here in the Philippines. My trainer there tells us that don't say the clause on Statistical technique is not applicable. Any evidence that shows the company is doing what it can to analyse data should be part of this element. But our Australian consultant tells us Statistics yes, but Statistical Analysis is an altogether different thing, ergo, NA. I've had at least 2 assessors saying no, and 1 says yes. Your site offers advice to not include SPC if its not applicable.

We're in the convenience store industry.


________________________
ahsan

ahsan,

This paper may inspire the company to obtain more value from statistical techniques instead of relying on legalistic interpretations of the standard:

http://www.managementjournal.info/download1.php?f=1202062013.pdf

Best wishes,

John

 

[Steve Prevette]

Re: When are Statistical techniques not appplicable?

The problem becomes acute in smaller companies where they do not have any personnel who are good with statistics. Many of them don't have anyone who is even good with Excel, which I consider a basic, even for data entry. I have a friend who ran a group of restaurants. They're gone now, but I remember not so many years ago he had a hard time finding managers for each "store" who could reliably input daily data into a pre-made Excel spreadsheet. I will say, in retrospect, that is, in my opinion, part of why his business venture eventually failed. He didn't keep his eyes on his data in many areas. And I told him so way back around 2005 before things started going down hill (2008ish things started to go "bad"). I had invested in his business by way of some loans. I got him to pay me off and got out or it (essentially I was an investor) back in 2005-6. I wouldn't even consider someone who isn't at least minimally "Excel knowledgeable".

I have often mulled over a business model of supplying centralized statistical services, and I know some folks here have tried to do that. It is unfortunate that there is not a market for the service, but a combination of "we don't need no steenking statistics" and mistrust of sending data "offsite" I think prevents this. Also, it is less than ideal to try to analyze data from a place you've never been.

I will say that within Fluor, I get sent to South Carolina the injury, event, and corrective action datafiles (in Excel) from our contract in Portsmouth OH each month, and end up charging them 4.7 hours a month (that is the usual time it takes, unless there is a data glitch) to make 362 charts. They find it to be a good deal, and has gotten good acculades from the government (DOE) client.

 

[Statistical Steven]

Re: When are Statistical techniques not appplicable?

I have often mulled over a business model of supplying centralized statistical services, and I know some folks here have tried to do that. It is unfortunate that there is not a market for the service, but a combination of "we don't need no steenking statistics" and mistrust of sending data "offsite" I think prevents this. Also, it is less than ideal to try to analyze data from a place you've never been.

I will say that within Fluor, I get sent to South Carolina the injury, event, and corrective action datafiles (in Excel) from our contract in Portsmouth OH each month, and end up charging them 4.7 hours a month (that is the usual time it takes, unless there is a data glitch) to make 362 charts. They find it to be a good deal, and has gotten good acculades from the government (DOE) client.

There is a market for these services. The problem is that it takes a compliance issue or some catastrophic event. But the biggest issue is finding enough recurring business to support a business. Assume a 160 hour month, 5 hrs to make 362 charts is 3% of the workload.....so you need 30+ clients.....not easy to do!