That's a possible outcome. You have to keep in mind that SPM's analysis is a multiple regression, and the significance of a regressor is computed in context of the other regressors. Essentially it is answering the question: "Does this regressor explain significant portions of the variance, ABOVE and BEYOND what the other regressors are explaining?". I would expect that you get similar results to SPM if you included both of your regressors in an external multiple regression analysis.
One thing I noticed is that your two regressors appear to be highly correlated. This may also affect the calculation of statistical significance.
Best,
Andreas
===================================================
Andreas Finkelmeyer, Ph.D.
Academic Psychiatry, Institute of Neuroscience
Wolfson Research Centre, Campus for Ageing and Vitality
Newcastle University
Newcastle upon Tyne
NE4 5PL, UK
Tel.: +44 (0)191 208 1357
Web: www.ncl.ac.uk/ion
>-----Original Message-----
>From: SPM (Statistical Parametric Mapping) [mailto:[log in to unmask]]
>On Behalf Of Roger
>Sent: 30 September 2013 18:36
>To: [log in to unmask]
>Subject: [SPM] Regression Values Question
>
>Dear fellow SPMers,
>
>I am running a second-level regression model to see if BOLD signal elicited by
>graphic images correlates with either or 2 behavioral measures (1 = ratings of
>the images emotional value, and 2 = effectiveness in dissuading smoking
>behavior).
>
>After running a conservative analysis (p < 0.005 unc), the signal in the
>amygdala appears to be significantly correlated with the responses to the
>second question (see attached).
>
>However, when I copy the Y-values out of the Matlab window and run a
>correlation between them and the responses by hand, it is not significant. I
>am at a loss to explain the difference between SPMs results and those I
>obtain running the regression by hand and hope you may have some advice!
>
>Roger
|