Fix get_e1g2T about ngmix HOT 5 CLOSED

Should the row col be relative to the center?

row,col=self.get_cen()
rowdiff=gm['row']-row
coldiff=gm['col']-col

irr=((gm['irr'] + rowdiff**2)*gm['p']).sum()*ipsum
irc=((gm['irc'] + rowdiff*coldiff)*gm['p']).sum()*ipsum
icc=((gm['icc'] + coldiff**2)*gm['p']).sum()*ipsum

from ngmix.

Yes. Sorry, I just assumed (always a mistake) that that was what the 'row', 'col' parameters were. They should be the r,c offsets relative to the centroid.

This formula is basically just the parallel axis theorem for second moments. I_offset = I_center + r^2.

from ngmix.

Yes, thank you, I'll implement this soon.

from ngmix.

The parallel axis theorem seems like a sledgehammer for these purposes! However, now I can say I have seen it used somewhere in real research, as opposed to before when I just needed it to solve made up mechanics problems.

from ngmix.

I remember someone (maybe Mike even) suggesting this solution to me in the past, but I quickly forgot it and didn't implement it.

I've pushed a change to master that implements this.

from ngmix.