Statistics / Excel Help?

[keith]

Well, I thought I had DataPage on the computer, but apparently it's not licensed. I get a missing authorization code error.

At any rate, I measured 6 identical parts (1222 points on each part). I exported the data to a CSV and imported the data into Excel, one worksheet per part. I wrote a macro to get just the Deviation (T) for each part on one worksheet. My data looks like:

Point Part1 Part2 Part3 Part4 Part5 Part6
1 value value value value value value
2 value value value value value value
3 value value value value value value
4 value value value value value value
...
1219 value value value value value value
1220 value value value value value value
1221 value value value value value value
1222 value value value value value value

The values are the actual deviations at each point for each part.

What kind of statistics can this data show? I can find Max and Min for each part and for each point, but I'm not sure what that tells me. I guess what I'm saying is that I don't know what to do statistically to show something meaningful. How can I show which part is closest to CAD? Compare Mean value? Average values? Create a histogram?

I have the Data Analysis Pack installed, if there is something in there I that would help.

Thanks in advance,
Keith

[craiger_ny]

Be careful doing statistics with Excel. Standard deviation is not the same in Excel as it should be calculated in many cases. As a matter of fact standard deviation is a dynamic expression that is calculated differently (usually an estimated figure) depending on the type of data and the type of statistical test you are running. You can get close but there will be a difference from your numbers and someone else's who obtains the standard deviation another (perhaps the proper) way. That difference can add up when standard deviation becomes a part of numerous equations. At any rate you may need to give us more info as what you want to do is doable ignoring weather or not it is statistically sound. What are you hoping to learn?

Craig

[T Miller]

You should be able to get your Average,and range with your min max and if you have your noms the differance from nom.
=average range is the differance from min max

[keith]

Thanks for the quick response.

We ran a part on virtually 6 different machines. We want to know which machine will produce the most accurate part. I printed the graphics window for each part and can visually see which part looks best, but I'd like to show some sort of numerical analysis, too.

[keith]

Thanks for the quick response.

So I could find the min + max for each part, then get the range (max-min). That would show me the accuracy of the machine?

The values I have in Excel are X,Y,Z,T. I copied the T values into one worksheet. Wouldn't those be the difference from nom?

[Matthew D. Hoedeman]

There are 2 ways to calculate STD-DEV, ESTIMATED and SAMPLE. Excel has 2 primary STDEV calcualtions, STDEV and STDEVP. STDEV uses SAMPLE calculation while STDEVP uses ESTIMATED for it's calculation. Datapage can be set up to use ESTIMATED STD-DEV for Cp and Cpk calcualtions and SAMPLE STD-DEV for Pp/Ppk calculations. OR, you can set Datapage to use SAMPLE STD-DEV for Cp/Cpk and no Pp/Ppk. Pretty much, it (should) be set to use the first option, ESTIMATED of Cp/Cpk and SAMPLE for Pp/Ppk. So, to use Excel, you need to know, WHAT are you trying to calculate? A study based on SAMPLE or ESTIMATED (sample measn 100% are checked, estimated is for 10 out of 1000 or whatever get checked). Once you figure out what you are checking you can correctly set up Excel to calculate your values (Pp/Ppk ~ Cp/Cpk).

[keith]

Thanks for the quick response.

So if I calculate the STDEV of all the points for each part, I get a number for each part. 1 is .003149, 2 is .00178986, 3 is .00380761, 4 is .001632143, 5 is .003225561, and 6 is .001524. What do these numbers represent? How do I interpret the STDEV?

[craiger_ny]

Standard deviation is a measure of spread or process dispursion. the higher the number the more dispursion. More = bad, less = good. Do you have access to any statistical manuals? You are on line so that is a start. It sounds like you are starting from zero here. Be careful.

Craig

[keith]

Thanks for your correct assessment here. I had a little statistical training a long time ago (> 15 yrs).

I have just realized another question, though. If T is the deviation from the CAD data, why are some numbers + and some - ?

[Matthew D. Hoedeman]

Are you calculating the STD DEV for the 1220 points of part#1 and for the 1220 points of part #2, etc.? Is that what you are saying? That is not really the way it should be used, you should be getting the STD DEV of POINT #1 for the 6 parts. However, it sounds like you want to find which part of the 6 checks the best out of the 6, right? If so, Statistics isn't really the way to go, you pretty much have to look at each point of each part. If you are trying to find the best part of the 6, the over-all average of all the deviations of each individual point can work and can be used, as long as you also take into account the MIN and MAX readings from each part. A better method might be just to look at the MIN and MAX and the RANGE of all the point from each part (and this is something that Datapage CAN NOT DO, but Excel can). It does you no good to say Part #1 has the best average is ALL the points check -0.0002" except for one point which checks +0.244". With 1220 points, that would give an average of ZERO (PERFECT, eh?) while another part has an average that is slightly off of zero but ranges from -.0005" to +0.0004". So, if you are looking to see which part is best, you would need to look at:

1) MAXIMUM reading of each part
2) MINIMUM reading of each part
3) RANGE of each part
4) AVERAGE of each part

Personally, I would probably say the part with the smallest RANGE is the best one, AS LONG AS THE PROCESS can be repeated. If it can be repeated, then a simple adjustment SOMEWHERE (cutter comp, tooling choice, machine pick-up, etc.) will tweak it until it also has the best average deviation.

[Matthew D. Hoedeman]

A PLUS (+) deviation means that the part is HEAVY (or + stock) while a MINUS (-) deviation means the part is light (too little stock). OR, another way of looking at it, PLUS means the probe made contact too soon and minus means it made contact too late, or +=towards probe start position, -=away from probe start position.

[keith]

thanks, that what my boss and i were just assuming

[keith]

thanks, Matthew

[P Roberts]

If you are looking for a way to compare the relative volumetric accuracy of the 6 parts you should take the avereage of the absolute value of the variances of each point, this would give you the averaged deviation of each part. For example create an extra colum for each part that takes absolute value of each point (=abs(B1), filled down for each 1220 points) and take an average of these absolute values for each part. (=average(C1:C1220)).

[spazus_maximus]

Matt you ruined my morning! I thought estimated stdev was used for Pp Ppk & Sample stdev for Cp & Cpk. My capability templates are wrong! Please dont tell the process guys.