Maybe I misunderstood Sabine Hossenfelder but she wrote:

> They just throw away data that don’t look like they want it to look.

If I take white noise and an assumption about how my data should look like I can basically prove anything.

If I take two white noise generators I am pretty sure I will still be able to find a signal which more less appears at the same time at both generators.

I hope they did more than just select the data which fits their assumption.