When to use Intraclass Correlation Coefficient vs. Coefficient of Variance

Elsmar Forum Sponsor

Miner

Forum Moderator
Staff member
Admin
#12
You may not get a response over the weekend because the forums are usually slow at that time. I'll take a look at your data on Monday.

I recommend that you also plot your data in time sequence and evaluate it for any time-related trends.
 

Miner

Forum Moderator
Staff member
Admin
#14
Nevertheless, you should always analyze for time trends. If you see trends that should not be there, it is an indication that an unknown factor is at work.
 

Bev D

Heretical Statistician
Staff member
Super Moderator
#15
I'll probably look at the data today but the first comment I will make is that you are better off measuring 20-30 parts (or outcomes as you call them) than having each person measure the same part more than 2 times.

when you are trying to estimate measurement error you are estimating a standard deviation. estimates of the standard are more accurate with either a relatively large single sample (30-60 pieces if sampling a population) or even better is 30 sets of small samples over time. When we apply this to measurement systems the best approach is many parts with 2 measurements for each part/person.

As stated earlier CV and ICC are NOT statistics of the same type of variation. CV will help when there is an interaction between the measuing system and the size or property of the thing being measured. in other words the variation at the low end is different than at the high end. This typically happens when measuring a biological or chemical properties as the concentration varies...it is rarer in mechanical or electrical features or properties.
 

Bev D

Heretical Statistician
Staff member
Super Moderator
#16
One question: what is the normal range of variation of your parts? or the spec limits.

from the data you posted you don't have a CV issue. the measurement variation isn't differetn for the different values of these parts. (IF I plotted them correctly)

you have a LOT of measurement variation for each putcome...the applicable statistic is ICC. but its' so bad you need to concentrate on improving your measurement system not on which statistic to use.

If you could it woudl also be helpful if you described the thing you are measuring, the measurement system adn how you are setting up the study.
statistics without physics is just an exercise in mathematics. at this point you need advice on physics...
 

Attachments

S

seirawan

#17
If we do not have a CV issue, shouldn't CV and ICC be very close to each other, why are they so different?
Please treat the measurments of each outcome as an independent study. The sample is composed of homogenous subjects.
 

Bev D

Heretical Statistician
Staff member
Super Moderator
#18
If we do not have a CV issue, shouldn't CV and ICC be very close to each other, why are they so different?
Please treat the measurments of each outcome as an independent study. The sample is composed of homogenous subjects.
because they are NOT the same statistic. They do NOT measure the same standard deviation...there is no reason for them to be the same...to compare them is like comparing apples and oranges...

The CV reports the standard deviation in the Y dimension for any given X value as a percentage of the X.

ICC is calculated using the standard of the differences of repeated meaurements for any part. dimensionally think of this as being the standard deviation about a 45 degree line (or 1:1 line) for the 2 repeated readings. The ICC statistic gives the amount of 'repeatability' vs the amout of observed variation...

but at this point your measurement variation is awful (UNLESS the 4 parts you measured are representative of a VERY small part of your overall variation) I would suggest first cocnentrating on fixing that - then teh statisitcs will actually make more sense to you...
 

Bev D

Heretical Statistician
Staff member
Super Moderator
#19
I might also suggest that you obtain a copy of Dr. Wheeler's book "Evaluating the Measurement Process". and studying how to set up a measurement systems analysis.
 

Stijloor

Staff member
Super Moderator
#20
I might also suggest that you obtain a copy of Dr. Wheeler's book "Evaluating the Measurement Process". and studying how to set up a measurement systems analysis.
Excellent suggestion!

Here is a source for Dr. Donald Wheeler's book on Amazon.com.

Stijloor.
 
Thread starter Similar threads Forum Replies Date
Ron Rompen Surface Finish Correlation between Ra, Rz and Tp (bearing surface ratio) General Measurement Device and Calibration Topics 3
R Preliminary MSA and Sockets' correlation Gage R&R (GR&R) and MSA (Measurement Systems Analysis) 0
Q How should I analyze measurement correlation between me and customer? Gage R&R (GR&R) and MSA (Measurement Systems Analysis) 12
S Significance of regression, correlation coefficient, hypothesesi, confidence interval Using Minitab Software 8
triathlonx13 Leak test correlation - Bubble to Air Reliability Analysis - Predictions, Testing and Standards 8
Ron Rompen Best Practices in CMM Correlation Studies General Measurement Device and Calibration Topics 3
L Coefficient of correlation (r) and b1 Statistical Analysis Tools, Techniques and SPC 1
Ajit Basrur MS Excel to study Correlation between Humidity & Differential Pressure Excel .xls Spreadsheet Templates and Tools 7
T Automated and Manual Measurement System Comparison (Correlation) Quality Tools, Improvement and Analysis 3
Q Correlation between ASTM & China's GB (Guobiao) / YY Standards China Medical Device Regulations 4
L ISO 9001:2015 Correlation Matrix & Change Summary ISO 9000, ISO 9001, and ISO 9004 Quality Management Systems Standards 3
M How to find the number of required data - Correlation with 4 variables Using Minitab Software 2
M Correlation matrix of ISO 13485 and other country regulations Other ISO and International Standards and European Regulations 3
Q Correlation Matrix of ISO 9001:2008 and ISO 9001: Wanted ISO 9000, ISO 9001, and ISO 9004 Quality Management Systems Standards 2
M Is correlation matrix required for Customer Specific Requirements in Quality Manual? IATF 16949 - Automotive Quality Systems Standard 3
R Compression Spring Force measurement correlation with Supplier Capability, Accuracy and Stability - Processes, Machines, etc. 7
N Correlation vs. Causation - Why Correlation does not imply Causation Example Philosophy, Gurus, Innovation and Evolution 7
J Correlation between OEE and FMEA FMEA and Control Plans 2
W Zero setting of correlation graph when plotting best fit linear regression Statistical Analysis Tools, Techniques and SPC 7
A Statistical Correlation between ordered SKUs Statistical Analysis Tools, Techniques and SPC 8
C Test Correlation - Moving Test Equipment to a New Location Manufacturing and Related Processes 7
J Correlation matrix between HSA TS-01 and 21CFR820 or ISO13485 Other Medical Device Regulations World-Wide 1
P How to do Nonlinear Regression and generate a Correlation Statistical Analysis Tools, Techniques and SPC 14
B Correlation Coefficient and Industry Standards - Corrrelation between two of my CMMs Using Minitab Software 2
Q Correlation Matrix and Value Stream Mapping? Quality Assurance and Compliance Software Tools and Solutions 1
V Identifying Significant Factors - Regression Analysis vs Correlation vs ANOVA vs DOE Quality Tools, Improvement and Analysis 7
S The relationship (correlation) of 5 Core Tools with TS 16949 IATF 16949 - Automotive Quality Systems Standard 9
C What is difference between a t-test and Correlation Coefficient Statistical Analysis Tools, Techniques and SPC 2
optomist1 Correlation and Causation - Causation seems to be Ambiguous and Slippery Quality Tools, Improvement and Analysis 7
M Correlation between 2 Variables - Potential X's that might affect yield percentage Quality Tools, Improvement and Analysis 5
S Correlation Study between a new Measurement Device and an existing Device Gage R&R (GR&R) and MSA (Measurement Systems Analysis) 18
M Correlation Production Time Study on a Process on an Assembly Line Manufacturing and Related Processes 3
O Correlation of 2 Test Devices - 2 samples (13 parameters) Gage R&R (GR&R) and MSA (Measurement Systems Analysis) 6
K Definition Correlation - The meaning of correlation and and its application in data analysis Definitions, Acronyms, Abbreviations and Interpretations Listed Alphabetically 4
N Correlation of Gage R&R Value with the Measured Value of Parts Gage R&R (GR&R) and MSA (Measurement Systems Analysis) 6
Q ISO 13485, FDA's QSR, CMDR, MDD: Seeking Internal Audit Correlation Matrix Various Other Specifications, Standards, and related Requirements 3
D Improving coefficient correlation result Statistical Analysis Tools, Techniques and SPC 1
H Use SPC to Calculate Correlation Device Tolerance - Hybrid Electronic Components Capability, Accuracy and Stability - Processes, Machines, etc. 3
P Is there a correlation between Teen Violence & Video Games? Coffee Break and Water Cooler Discussions 28
D Two Gauges (Gages) - Two measurements correlation I do not know how to explain Reliability Analysis - Predictions, Testing and Standards 4
S Correlation or Mann Whitney test results? Which to use Statistical Analysis Tools, Techniques and SPC 1
V Correlation and determining the % the contribution Six Sigma 3
R Checking Correlation Between Two Similar Gages Measuring the Same Parts Gage R&R (GR&R) and MSA (Measurement Systems Analysis) 14
J Correlation using t-test and correlation coefficient - What is the correct result? Statistical Analysis Tools, Techniques and SPC 43
A Inter laboratory correlation - Two laboratories with different methods? General Measurement Device and Calibration Topics 4
SteelMaiden Competency evaluation records - Direct correlation to EMS-related topics Miscellaneous Environmental Standards and EMS Related Discussions 7
A Correlation or Hypothesis Test? Which is approprite to use on what kind of problem? Statistical Analysis Tools, Techniques and SPC 6
D Correlation of two tests that are NOT linearly correlated? Statistical Analysis Tools, Techniques and SPC 11
H Process correlation matrix - Just how far "into the weeds" do I go? ISO 13485:2016 - Medical Device Quality Management Systems 4
S MSA (measurement systems analysis) on several CMM's - Correlation issue Gage R&R (GR&R) and MSA (Measurement Systems Analysis) 3

Similar threads

Top Bottom