I am currently investigating the effect of x on y. In a single sort it is evident that the higher x is the lower y will be. If however, I first sort according to z and within the different z buckets I sort then again according to x, x has no effect on y.
However, if I do a cross-sectional regression: y ~ x + z both x and z have a significantly negative impact on y. Shouldn't the effect of x in such a case also be not significant?