Right correlation formula to use

Hi!

I have a 12x1 vector, and a value. I would like to find the correlation of each of the vector values with my variable. How can I do it?

I tried with corr, corrcoef, corr2 and I only got NaN, why is that?

To clarify my vector is

x = [0.0000827

0.000229

0.000358

0.000486

0.000561

0.000564

0.00483

133

167

178

213

386

];

and my value is y = 2.261133333.

Any suggestions?

Thank you,

Carlotta.

[Jos (10584)]

I think you should read a primer on statistics what a correlation actually is. A correlation is a measure of association between two series of related paired numbers (Xi, Yi), i=1:N

Maybe you mean something else by correlation? What kind of output did you expect?

[Carlotta Fabris]

Thank you for your answer Jos!

The data I have are coming from an analysis I did. I had some fMRI data and I did a GLM to obtain activation areas for some stimuli.

The vector I have represents the peak of activation for each active area. The value I would like to correlate these peak with is the mean of some dependent variables.

I was hoping to obtain as result if the DVs correlate with the peaks or not.

[Jeff Miller]

As Jos said, you are not using "correlate"/"correlation" in the way it is normally meant.

A correlation is only defined for a two-dimensional (i.e., Nx2) array, and there is no such thing as a correlation with a constant like 2.261133333.

It sounds like you might be trying to find some measure of similarity of some values to that constant (e.g., differences from it), but correlation is not the right measure, because there is no such thing for this case.

[Carlotta Fabris]

Thank you Jeff!

I understand and probably there is a better approach to my problem.

I have these peak values and some DVs, what I did was find the mean of the DVs I was interested with to correlate them. As you both said, this is not possible, so do you have some suggestions on how I should do it?

Thanks again,

Carlotta.

[Jos (10584)]

You have to define "it" more precisely in the sentence "how should I do it?", Carlotta. Most of us here are quite willing to help, but do not possess a crystal ball, alas. ~Jos

[Carlotta Fabris]

By it I mean find a correlation between the peak values and the DVs I have.

Thank you,

Carlotta.

[Image Analyst]

Maybe you just want the differences between y and x:

% Define DV "peak values":
x = [0.0000827
0.000229
0.000358
0.000486
0.000561
0.000564
0.00483
133
167
178
213
386
];
% and my value is 
y = 2.261133333;
indexes = 1 : length(x);
plot(indexes, x, 'b*-', 'LineWidth', 2);
hold on;
plot(indexes, y*ones(1, length(x)), 'r*-', 'LineWidth', 2);
grid on;
% Compute absolute differences
absDiffs = abs(x - y)
% Compute average of those
meanAbsDiff = mean(absDiffs)

absDiffs =

2.261050633

2.260904333

2.260775333

2.260647333

2.260572333

2.260569333

2.256303333

130.738866667

164.738866667

175.738866667

210.738866667

383.738866667

meanAbsDiff =

90.1262629971667

[Carlotta Fabris]

Thank you for your suggestion! I will consider this, but what if I don't do the mean of my DVs? What could I do having the peaks and 250 DVs each one having 23 values between 1 and 4?

I hope I was clear, if not, ask me!

[Image Analyst]

It is not clear. I'm not sure what DV is. Are the DV values the values in x? But x has only 12 continuous values, not 250. And only the fractional part seemed to be quantized into levels. And the value of x are not between 1 and 4. Please attach your data.

And I agree with the others in that it does not make sense to correlate a single value against a set of many other values - you can only correlate one set with another. Make another shot at trying to clearly explain things.

[Carlotta Fabris]

I am attaching you an Excel file with the DVs. What I talked about before was the mean I did of these DVs considering the ones I was interested in.

I have the peak values and these DVs, what I would like to understand is: is there a correlation between the peaks and the DVs? How can I calculate this correlation? If this is not possible, then what could I do?

Thank you.