Comment by robocat

Surely sampling bias would be really hard to remove from this.

How did they control for couples that remove themselves from the sample set because they broke up due to bickering (or a correlation with bickering)?