Calculate P values at a point below the critical threshold.

[ArjS21]

Hi, I have used a general linear model implemented using SPM and from the SPM(t) plot, there is one supra-threshold cluster, and another region where the curve approaches the critical threshold closely, but does not cross the threshold. I would like to calculate the P value at this point as a descriptive variable and refer to it when qualitatively describing my data. I have looked at the spmi variable and outputs but am unsure how I can calculate the P value at this region using the outputs available. Can this be done with the t* and z outputs?

Thanks, A

[0todd0000]

Yes, this can be done, but please note:

• SPM's p-values become less accurate as p increases when p>0.5
• Pointwise p-values are different from cluster-level p-values, so they shouldn't be compared directly.

That being said, assuming two-tailed inference, you can calculate pointwise p-values for the t-statistic like this in Python:

spm = spm1d.stats.glm(y, X, c)
p   = spm1d.rft1d.f.sf(spm.z**2, spm.df, spm.Q, spm.fwhm, withBonf=True)

I think that should work in MATLAB (after changing spm.z**2 to spm.z.^2 and using 'withBonf', true) but let me know if you're unsure.

[ArjS21]

Thanks Todd. I have changed the variables mentioned to MATLAB syntax, but get a matrix dimensions error, as below. I'm not too sure on where this is stemming from.

Matrix dimensions must agree.

Error in spm_P_RF (line 82) EM = (R./G).*P; % over D dimensions

Error in spm_P (line 51) [P,p,Em,En,EN] = spm_P_RF(c,k,Z,df,STAT,R,n);

Error in spm1d.rft1d.RFTProbability/upcrossing (line 23) p = spm_P(1, 0, u, c.df, c.STAT, c.resels, c.n, c.Q);

Error in spm1d.rft1d.RFTCalculator/sf (line 58) p = self.p.upcrossing(u);

Error in spm1d.rft1d.f.sf (line 38) p = calc.sf(u);

[0todd0000]

Apologies for the delay! This slipped through my inbox.

It seems that the MATLAB sf function cannot accept arrays so instead try calculating for each time point like this:

Q = spm.Q;
p = zeros(1,Q);
for q = 1:Q
    p(q) = spm1d.rft1d.f.sf(spm.z(q)^2, spm.df, Q, spm.fwhm);
end