In a comment reply on Frank Harrell's answer here, in relation to interpreting an interaction, he says this should be done by:
double difference contrasts and a series of single differences at different levels of interacting factor.
What exactly does this mean, and can someone please show an example of doing it ?
