Correlation using t-test and correlation coefficient - What is the correct result?

[jas4710]

In Reply to Parent Post by Bev D

Additionally I will reiterate as Steve did that paired t tests and linear regression are NOT the appropriate statistical tools for measurement studies such as yours. The Intraclass correlation coefficient is the appropriate statistic...a simple google search for this will help.

From SPSS, there are different parameters

(Two-Way Mixed when people effects are random and the item effects are fixed, select Two-Way Random when people effects and the item effects are random, or select One-Way Random when people effects are random.)

However, the so-called "people effects", "fixed", "random" look unclear to me. Furthermore, should only one set of data is to be supplied or both sets?

[prototyper]

At this point there are too many unknown factors!
We do not know the repeatability and reproducibility of either piece of measuring equipment, if there is any operator test retest error or operator bias.

I would recommend a study as per the MSA manual third edition on each individual piece of equipment to identify any weakness in the measurement systems prior to any direct comparison of data.

I have attached a spreadsheet which I developed from the manual which may help. The data included is directly from the MSA third edition to validate the calculations.

[jas4710]

In Reply to Parent Post by prototyper

At this point there are too many unknown factors!
We do not know the repeatability and reproducibility of either piece of measuring equipment, if there is any operator test retest error or operator bias.

I would recommend a study as per the MSA manual third edition on each individual piece of equipment to identify any weakness in the measurement systems prior to any direct comparison of data.

I have attached a spreadsheet which I developed from the manual which may help. The data included is directly from the MSA third edition to validate the calculations.

MSA is too rigorous for me to follow... is there any way to better perform intraclass correlation coefficient analysis? Merely taking all data will yield poor R but I just don't know how intraclass R can be calculated *meaningfully*.

[prototyper]

Not that I know of as the intraclass correlation coofficient is the ratio of the estimated product variation and the estimated measurement variation.
We do not know the measurement variation inherent from each piece of equipment.

[Bev D]

In Reply to Parent Post by jas4710

MSA is too rigorous for me to follow... is there any way to better perform intraclass correlation coefficient analysis? Merely taking all data will yield poor R but I just don't know how intraclass R can be calculated *meaningfully*.

what are you trying to do?

[jas4710]

In Reply to Parent Post by Bev D

what are you trying to do?

The same as the OP. I've got 2 sets of data in hand and discovered correlation of around 0.6. From studying the previous messages, it seems that intraclass correlation coefficient may help explain which portion of the data is leading to poor correlation. Therefore I'm asking how to calculate "intraclass correlation coefficient " Thanks a lot

[Bev D]

the OP had repeated measurement error data. (two independent measurements of 20 different samples)
Is this the type of data you have? or do you have measurements of two variables that you suspect as having a cause and effect relationship to each other?
also if you POST your data and explain what it is, how you collected it and why you are interested in it we can help you.

the intraclass correlation coefficient ahs a very specific applicability adn i don't know from your post if it applies.

[jas4710]

In Reply to Parent Post by Bev D

the OP had repeated measurement error data. (two independent measurements of 20 different samples)
Is this the type of data you have? or do you have measurements of two variables that you suspect as having a cause and effect relationship to each other?
also if you POST your data and explain what it is, how you collected it and why you are interested in it we can help you.

the intraclass correlation coefficient ahs a very specific applicability adn i don't know from your post if it applies.

Columns D & G are the data obtained by measuring the movement of different parts (Column A refers to 3D coordinate ID) of an object under different experimental conditions, e.g. different temperature. Maybe more than 1 condition is varied (not designed by me) and I'm just testing whether under different conditions shall still give readings with good correlation. If not, is particular part of that object leading to difference? Thanks once agin.