Most of the tools and published queries calculate a "relative" skew value rather than some sort of absolute measure.
If most AMPs have only table headers but a few AMPs also have a data row, the calculation will show a high relative skew. But in absolute terms, the difference is trivial (and expected).
Rather than trying to adjust the definition of "skew", I generally suggest people ignore skew until the total amount of data exceeds a few MB per AMP.
Most of the tools and published queries calculate a "relative" skew value rather than some sort of absolute measure.
If most AMPs have only table headers but a few AMPs also have a data row, the calculation will show a high relative skew. But in absolute terms, the difference is trivial (and expected).
Rather than trying to adjust the definition of "skew", I generally suggest people ignore skew until the total amount of data exceeds a few MB per AMP.