My response variable is a count (1, 2, 3 ... etc.) with right skewness, and actually we use the inverse of the count (1/count). So a count of 1 gives 1, 2 gives 0.5, 4 gives 0.25 and so on. The data distribution changes from right to left skewness.
Gamma regression is usually recommended with right skewed data. If not suitable for left skweness, what is the alternative?