Hello,

I work in a transactional area and was in the process of setting up a Gage R&R study. We pull queries of data from a system on a daily basics and that is the system I want to perform a Gage R&R for. Currently, the set up is as follows...4 queries pulled 2x by 5 operators. I want to determine if the output matches with the "master" query. The criteria will be good/bad. Am I on the right track? I plan to use Minitab to evaluate the data.

Thanks.

Hello,

I work in a transactional area and was in the process of setting up a Gage R&R study. We pull queries of data from a system on a daily basics and that is the system I want to perform a Gage R&R for. Currently, the set up is as follows...4 queries pulled 2x by 5 operators. I want to determine if the output matches with the "master" query. The criteria will be good/bad. Am I on the right track? I plan to use Minitab to evaluate the data.

Thanks.
Added in edit: Welcome to the Cove :bigwave:

Is this a Six Sigma question? Why do you want to perform GR&R when there doesn't appear to be a gage involved? It seems that some other form of statistical analysis would be more appropriate.

Hom er-

Can't tell you if this is standard practice in your line of work - my experience is in machine shop environments doing R&R on such things as micrometers, calipers, pretty much STME type stuff. I always did 10 samples with 3 operators doing 3 repeats for each of the 10 samples. The 3 operators gave me the reproducability, the 3 reps each gave me the repeatability. It also made for a really easy Excel sheet to record the results and calculate the R&R. If I only had 2 operators handy for the study, it was a simple matter of changing one constant in the sheet to compensate. When I get back to work after my nice long vacation, I'll post my Excel sheet here.

Regards -John

The important thing is we want to segregate the measurement variation out from the actual process variation when we observe the process variation.

IMO, even the visual criteria is not a gage, but we still can use the same technique to evaluate how effective on the visual criteria is. To do this we can use the attribute agreement analysis to evaluate agreement for between appraisal, within appraisal and to the stardard (visual criteria itself).

Standardize the method is important to improve the attribute agreement analysis and it is a usefull technique.:)

Welcome to the Cove,

As with the others who have already posted I'm a bit confused as to what you're looking for. A gage R&R is typically to monitor measurement variation of an instrument. Since you have no instrument involved it's not a suitable tool in this scenario.

Perhaps an attribute control chart would be more suitable for your task...if you could provide a bit more insight into what you want to accomplish we could likely offer a few useful tools/ideas/spreadsheets on the subject.

:D

Ok, here's a bit more information on what I'm tyring to accomplish. We have a web-based system from which we pull data. The concern is that the data may not be the same everytime its pulled from this web-based system and may not be the same when pulled by different people. Therefore, I wanted to set up a study where different people would pull a set of data using the same serach criteria at different times from the web-based system. Basically, are the results the same between people and are the results the same within each person when they pull the data more than once. Also, how accruate are the to the "reference" data set. I hope that helps. I appreciate all your comments.

Thank you.

We have a web-based system from which we pull data. The concern is that the data may not be the same everytime its pulled from this web-based system...
Why would it change? Don't you have to verify that it does change (or doesn't) before going any further?

I'm curious as well, why would the data not be the same if it's being pulled from a database.

Yes I suppose you could pull all that data together and evaluate it in minitab as a form of R&R but the question arises, to what purpose. Will it be valuable to you to see who has the most variation? If they truly are pulling it from a static database source, then the only reason it would change is if they are manipulating it in some way after retrieving it....truncating or rounding differently?

If it were me I'd just document who screwed up, how they screwed up and what percentage of the time they screwed up. The r&r will give you a good bit of information that you don't seem to need, though it may look 'neato' to the boss when you present it.

If you need the variation info user to user and such, then yes I'd say go ahead with a r&r, perhaps the different users are using different rounding methods and you want to show the variation it's causing in your records? Anyhow, if you don't need that info, I'd go with something simpler.

I'm curious as well, why would the data not be the same if it's being pulled from a database.

Well, if the queries are written by different people they will work differently. especially if there is any categorization and counting of free form text.
If the databases have "redundant" data in different areas...
and of course one could do a R&R on the INPUT of the data - especially if data is judged such as that doen to categorize defects or complaints...

Yes I suppose you could pull all that data together and evaluate it in minitab as a form of R&R but the question arises, to what purpose. Will it be valuable to you to see who has the most variation? If they truly are pulling it from a static database source, then the only reason it would change is if they are manipulating it in some way after retrieving it....truncating or rounding differently?
you never know until you know...if he -or others are truly unsure then they shoudl verify it...

If it were me I'd just document who screwed up, how they screwed up and what percentage of the time they screwed up.

"screwed up"? I guess you haven't been following the threads on operator error and their causes...

Well, if the queries are written by different people they will work differently. especially if there is any categorization and counting of free form text.
If the databases have "redundant" data in different areas...
and of course one could do a R&R on the INPUT of the data - especially if data is judged such as that doen to categorize defects or complaints...


Until it's known that the same queries will return the same responses every time (if that's what's supposed to be happening), there's no point in looking at what operators are doing. Some part of this has to be controlled in order for any type of experiment to yield useful results.

Until it's known that the same queries will return the same responses every time (if that's what's supposed to be happening), there's no point in looking at what operators are doing. Some part of this has to be controlled in order for any type of experiment to yield useful results.

it all depends on the data, the databases and what a priori knowledge you have. I have started several projects with the 'operator entry' side since I knew their was considerable judgment to the categories and so if the categorization was unreliable it didn't matter if the queries were reliable...
on the other hand I've also had projects where the database was redundant and queries were 'ad hoc' and 'known' to not match but the data was not judgment based and so was more likely to be relatively reliable. In these cases I started at the query end...

it all depends.

it all depends on the data, the databases and what a priori knowledge you have.

Which was my point, at least indirectly. Nonetheless, there still must be an element of control somewhere. If we know that a properly-formed query will return consistent results, then we can look at the effect of operators. On the other hand, if we know that operators are well-trained in constructing queries we can test for the quality of the database and/or the programming. One or the other must be controlled before results will make any sense. We need to identify the null hypothesis, in other words.

"screwed up"? I guess you haven't been following the threads on operator error and their causes...

I like the referral to operator error threads :lol: you're right, it's process error;)

The sentence was just my way of being facetious and saying I would try and keep it simple. It would seem that the areas where variation could occur are relatively limited and so I would attempt to locate where that variation is coming from, or if it exists, before trying to measure how much variation there is between users.

Thank you all.
Maybe a Gage R&R isn't the appropriate word for the study. Based on the study set up, I hope to find 2 things...1) can I consistenly get the same results from the database and 2) are the operators well-trained to create queires and pull data.

Which was my point, at least indirectly. Nonetheless, there still must be an element of control somewhere. If we know that a properly-formed query will return consistent results, then we can look at the effect of operators. On the other hand, if we know that operators are well-trained in constructing queries we can test for the quality of the database and/or the programming. One or the other must be controlled before results will make any sense. We need to identify the null hypothesis, in other words.

controlling one of the factors, hmmm. but doesn't a properly designed R&R study provide you information on both of these separately and combined just liek any full factorial experiment? Aren't we as a profession trying to get away from holding factors constant and varying one thing at a time? the challenge as always to design the factors and levels appropriately but *most* R&R studies are pretty easy to design that way.

We need to identify the null hypothesis, in other words.

For the non-statisticians....http://en.wikipedia.org/wiki/Null_hypothesis

Stijloor.

We need to identify the null hypothesis, in other words.

For the non-statisticians....http://en.wikipedia.org/wiki/Null_hypothesis

Stijloor.

For the non-statisticians....http://en.wikipedia.org/wiki/Null_hypothesis

Stijloor.

Bad link, Stilljoor. Try Null hypothesis

Bad link, Stijloor. Try Null hypothesis

Jim,

I had the correct link at first.......don't know what happened.
Must be bad statistics.....or a bad Monday.

Stijloor.

Jim,

I had the correct link at first.......don't know what happened.
Must be bad statistics.....or a bad Monday.

Stijloor.

Demons. :D