I suppose I come forward with my own usage stats, but it is anecdata :)

And the andecdata matches other anecdata.

Maybe I'm missing why that's selection bias.