Lin's Correspondence coefficient (r)

[#1]

Hello,

I am trying to recreate an analysis that used Lin's Correspondence coefficient (r). The data are ordinal, not continuous.  I cannot find this analysis in JMP, are there any other suggestions.

More specifically, I have two variables and needing to see how they relate based on three different treatments. Simple regression is not suitable here since X and Y are not related. What is the appropriate analysis to do with this data? See JMP graph below.

Thanks,

Natalie

[#2]

Try Analyze > Fit Y by X. It should launch the Bivariate platform when X and Y are continuous. Click the red triangle at the top and select Fit Orthogonal > Univariate Variances. You will see a report like this:

See JMP Help for more information about the report.

[#3]

A script illustrates how the CCC is calculated and added to the Bivariate report layer. Note that line 3 would be omitted, and the resulting script would be run after the real data table is opened. The X role is for the standard method and the Y role is for the test method.

Names Default To Here( 1 );

dt = Open( "$SAMPLE_DATA/Method Comparison.jmp" );

dialog = New Window( "Launch Bivariate",
	<<Modal, 
	// require the user to select two variables before clicking OK
	<<On Validate(
		Show( xvar, yvar );
		If( Is Missing( xvar ) | Is Missing( yvar ), 
		// if xvar or yvar are missing do nothing when OK is clicked
			0
		,
			1
		);
	),
	Text Box( " Select two numeric columns. " ),
	H List Box(
		Text Box( " X, Factor " ),
		x = Col List Box(
			dt, // data table reference
			all, // display all columns from the data table
			// get the name of the selected column before the window closes
			xvar = (x << Get Selected)[1];
			Show( xvar );
		),
		Text Box( "Y, Response" ),
		y = Col List Box(
			dt,
			all,
			yvar = (y << Get Selected)[1];
			Show( yvar );
		)
	)
);
If( dialog["Button"] == 1, // if the user clicks OK...
	xcol = Column( dt, xvar ); // get the columns
	ycol = Column( dt, yvar );
);

obj = Bivariate(
	Y( ycol ),
	X( xcol ),
	Density Ellipse( 0.95, {Line Color( {66, 112, 221} )} ),
	Fit Line( {Line Color( {212, 73, 88} )} ),
	Fit Orthogonal( Univariate Variances, {Line Color( {61, 174, 70} )} )
);

rpt = obj << Report;

// Pearson correlation r three ways
regrC = rpt["Bivariate Normal Ellipse P=0.950"][Number Col Box( "Correlation" )][1];
regrR = Sqrt( rpt["Summary of Fit"][Number Col Box( 1 )][1] );
ob = (rpt << XPath( "//OutlineBox[contains( text(), 'Orthogonal Fit Ratio')]" ))[1];
regrO = ob[Number Col Box( "Correlation" )][1];

// Lin's concordance correlation coefficient
muX = Col Mean( :Standard );
muY = Col Mean( :Method 1 );
sigmaX = Col Std Dev( :Standard );
sigmaY = Col Std Dev( :Method 1 );
ccc = regrC * (2 / ((((muY - muX) ^ 2) / (sigmaY * sigmaX)) + (sigmaY / sigmaX) + (sigmaX / sigmaY)));

Show( regrC, regrR, regrO, ccc );

ob[TableBox(1)] << Append( Number Col Box( "Lin's CCC", { ccc } ) );

See NCSS source.

[#4]

Thank you!

Is the first solution Lin's correspondence analysis or is it called something else?  If I use the first solution, the univariate analysis, I want to be able to describe what I did in the methods correctly. Do you have any examples of the use of this analysis?

Thanks again!

[#5]

No. I confirmed that it is Pearson's correlation coefficient. I am not sure which analysis you refer to in your last reply. Can you repeat the question with specific references to the analysis about which you want to know?