Data Analysis Methods - No clue how to analyze this data

[faceless]

Hi,

first of all, I'm new and I'm in desperate need of help

I know nothing of statistics... Yet I have to analyze some data, and I guess I'm lucky I have MINITAB at my hands

I have to compare measurement methods

I'm measuring A with method 1 and 2
B with method 3 and 4
C with method 5
(where A and C should add up to B)

The results are numerical between ~0.1 and ~2.0, there are 10 samples with 1 to 6 results with each method.


For starters I'd be happy with comparing 1 and 2, and 3 and 4.
From what I gather, is an ANOVA the right tool to do this?

MINITABS Help files are immensely helpful, but I'm pretty clueless. The best programm won't help if you have no clue....



I hope someone could at least nudge me in the right direction...

thx in advance!

 

[bobdoering]

What is the question you are trying to answer? Which is the best gaging system? If that is the case, I would do a gage R&R on each, and pick the one with the best results. If that is not the question, then I guess some clarification of the issue would be good.

 

[faceless]

I'm not sure what that means.. so let me elaborate - I'm sorry I didnt make it clearer.

The (to me) best outcome would be if in the end I could say
- "Measurement system 1 is just as good as 2, 3 is just as good as 4."



- "To get A it is not necessary to do time intensive measurements with 1 or 2. Instead much faster measurements with either 3 or 4 may be done complemented by a measurement with 5."

For example.

Method 1 takes 12 hours and gives the Result 1,9

I'd much rather do
Method 4 taking 20minutes, giving 1,7
Method 5 taking 20minutes, giving 0,2
1,7 + 0,2 = 1,9

Same result much faster.



I hope this clears things up, thx for your help!

 

[Bev D]

sounds like you are doing (measurement) method comparison. this is a bit more complicated than simple Gage R&R.

HOWEVER, you really need to describe your situation in real words and not alpha numeric code.

Also it's important to understand if you already have data or or asking what data you need to collect.

we really do want to help...

 

[faceless]

I'm doing elemental analysis.

I'm measuring sulfate, sulfide and sulfur content of binder samples.
Where sulfate + sulfide should add up to sulfur. (Results are always given as weight percent Sulfur for easy comparison.)
Samples are destroyed upon measuring!


All the stuff on Measurement system analyis I found in the minitab help was very "automotive specific" and I'm not sure how to apply it.


From my data, which I already gathered (which probably was'nt the smartest move looking back...), I'd like to know:

Does Method 1 and 2 give the same sulfate amount ?
Because 2 only takes 20 minutes whereas 1 takes 12 hours!

Does Method 3 and 4 give the same sulfur amount ?
Because 4 takes only 20 minutes whereas 3 takes 1-2 hours!

Does the data fit the requirement that sulfate + sulfide = sulfur ?

Is it sufficient to measure Sulfur with Method 4 complemented by measuring sulfate with 2 or sulfide with 5 when dealing with an unknown sample.
Again because full measurement with all methods takes too long.



So you see this is not totally high end stuff.
It would be nice to be able to falsify or prove these statements, backed up by statistics. But if that doesnt work out I'll settle with throwing in some boxplots and interpretating wildly =)


thx!

 

[faceless]

More info needed?

every piece of advice will be appreciated, thx!

 

[wang1568]

The problem is trying to find out different measurement system which can be alternating each other, Please use the regression analysis to check out whether if the regression model is OK in your working region.

For example, x will be the wt % of content measured by method 1, and y be the wt % by method 2, if y=f(x) model is OK, then method 1 and 2 are alternative
The levels of x should cover the whole working region

 

[faceless]

Hi,

my data is not balanced. I have done 3 to 6 measurements on each sample with each method.

How should I go about this? Which measurements to leave out?!
Or should I just look at means?

 

[Miner]

If you post your data, we will be able to provide direct help to you. So far, we are essentially guessing because we do not fully understand your situation.

 

[Bev D]

Hi,

my data is not balanced. I have done 3 to 6 measurements on each sample with each method.

How should I go about this? Which measurements to leave out?!
Or should I just look at means?

please post this data in a spreadsheet with clearly labeled columns as to sample number and method number...

what you are attempting to do is "method comparison".
it is critical that your test each 'sample' with both measurement methods. This is known as PAIRED data. This is the essential part of thsi type of analysis.

Your samples should span the range of the expected variaiton in the element being tested...and please also provide the specification limits for each element.

at this point don't worry about sample size or anything else besides getting some data so we can show you how to to do the analysis...