@ThatAkhilRao @cjwich @BenHanowell May be these two? https://t.co/IeXaqjRTrL https://t.co/v4UucbAu1i
388 followers
8,756 followers
@Whoudini4ever @JohnMullahy @KhoaVuUmn It's arbitrary and result may not be invariant to the constant used. IHS is preferred by most. But it also sensitive to units. You can also use PPML (use a poisson and assume E[y|x] = exp(xb) rather than taking logs.
5,705 followers
@arthur_alb1 In the case of that last one though, proceed with caution https://t.co/f6GCYbmghT
1,375 followers
Yes, the scaling issue for IHS is extremely important, as the Aihounton and Henningsen article explains.