Nilearn 'effect_size' beta map yields very very high values (>1000)

Hi,

After working for a while with nilearn, mostly with '‘z_score’ maps, I recently looked into the actual values of my beta maps using ‘output_type’ as ‘effect_size’. I discovered my beta values are extremely high. A lot of them are above a hundred and even above a thousand, which does not make sense to me.
When doing second-level analysis and looking at the ‘z_score’ maps, everything seems fine, so I don’t know what is wrong.
I would appreciate any help! Thanks!
Here is my code for creating the beta map and the design matrix:

fmri_glm = FirstLevelModel(t_r=2,
                           slice_time_ref = 0.5,
                           mask_img = no_vent_bin_img,
                           noise_model='ar1',
                           hrf_model='spm',
                           drift_model='cosine',
                           high_pass=1./180,
                           signal_scaling=False,
                           minimize_memory=False)
        
test_glm = fmri_glm.fit(subj_imgs, subj_events, subj_conf) 
effect_map = test_glm.compute_contrast('feedback', output_type='effect_size')
plot_design_matrix(test_glm.design_matrices_[0])

Beta values are dependent on the scale of the BOLD data. Since you didn’t scale the data in your FirstLevelModel, you can check the scale of your input data to see if the beta values make sense.

Indeed, I would expect data scaling to be the culprit here.
Best,
Bertrand