1

I am investigating a moderator relationship after a significant interaction between two centred variables. Using simple slopes analysis I want to see the effects of high/low scores of one variable -1/+1 SD relative to the mean on another predictor. I thought I knew how to do this - but I am double-checking because a YouTube video just contradicted what I thought was a common-sense process.

I intend to add 1SD to every score (already centred on the mean) for the high-scoring group and subtract 1SD to every score (already centred on the mean)for the low-scoring group. The YouTube video advised that the opposite process should occur and that by adding 1SD to every centred score you would have the low-scoring group and vice versa. Is this true?

e.g.

(variable - m + 1SD = low scoring group) & (variable - m - 1SD = high scoring group)

https://youtu.be/sY8lJ3XJ5JM?t=195

Please see the above YT link timestamped at the relevant moment.

Max Moser
  • 11
  • Link to the video for context? You are right on what you describe. Maybe there is an explanation on why the person said that. – POC Mar 16 '21 at 18:22
  • Please see edited OP - I have timestamped the video to the relevant moment. Thank you. – Max Moser Mar 16 '21 at 18:27
  • This is certainly not how I am used to carry this analysis and you probably thinking of another technique also. You probably think about using the original $\beta$, and not those recomputed $\beta$ like he does. – POC Mar 16 '21 at 18:56
  • Do you think I should be using another analysis to explore an interaction? I am from social sciences and am very new to statistics. – Max Moser Mar 16 '21 at 19:01
  • II would recommend the Johnson Neyman technique https://www.tandfonline.com/doi/full/10.1080/10705511.2020.1732826 or you can find many techniques in Field, Howell or other book on statistics in the beahvorial sciences. – POC Mar 16 '21 at 19:16

0 Answers0