Sept 1st, 2019 - Lenspix Inducing Negative Pixels?

When I took the log of the lensed CIB in the previous post, an error turned up telling me that some values were invalid. So I took a look at the histogram (below) and the lensed CIB certainly has some negative values (about 0.1% of all the pixels). It seems to me that the cut at 0 for the unlensed CIB is abrupt, and when Lenspix simulates the lensing, it tends to smooth out the histogram, without taking into account that some values are un-physical. I took a look at other lensed maps, including the null-lensed maps and they seem to have negative values as well. The question is, how much do these negative-value pixels affect our non-Gaussianity analysis? From a previous post, (https://mocks.cita.utoronto.ca/index.php/July_19th,_2019_-_Reduced_Bispectrum,_0th_Shell_Additional_Plots), I show that the interpolation effect by Lenspix (including the negative-value pixel effect) is tiny compared to the change in the bispectrum due to lensing. However, I have not calculated the interpolation effect after masking the brightest pixels (which are not affected by the lensing at all) although this probably will not cause a big difference. Aspects of non-Gaussianity other than the reduced bispectrum in the equilateral configuration may be affected more, so we should be aware.

It seems that we (I) should take more care in validating our results when using 'lenspix' (being aware of what limits it has).