I have written a paper in which I run a 2x2x3 repeated measures ANOVA using SPSS. My dependent variable is reaction time and my repeated IVs are different visual conditions. I originally also had a between-subjects variable which was "high" or "low" symptomatology (based on a median split of scores on a measure of trauma). A reviewer has suggested that the trauma measure should be entered as a continuous predictor rather than grouped. They suggested that I enter this continuous predictor as a covariate in the ANOVA. I have done this as I do have a significant interaction between trauma score and one of my main effects ('face type') which I interpret means that the effect of 'face type' on reaction time is moderated by trauma score. What is the best way of following this up further? My reviewer suggested using the "Aiken and West" procedure of entering values one standard deviation above and below the mean, but when I have looked into this it seems to be for regressions than for ANCOVAs. I have performed a correlation for the relationship between trauma score and RT under the different face type conditions, but this isn't really what I'm interested in because I'm more interested in how trauma influences the DIFFERENCE in RT between the different face types. Furthermore, when I run the correlations for trauma score +/- the standard deviation, the r is the same (which I've just realised is obvious as all I'm doing is transforming the values). If anyone can advise or point me in right direction I would be forever grateful! Many thanks
it sounds like you are interested in the interaction (in the sense of linear models) between trauma and face types - adding this interaction in your model should give you an answer.
Dear Chloe,
If you have only repeated the experiment once you could reformulate it as a multiple linear regression (MLR). ANCOVA does not allow for different slopes of regression lines whereas MLR does. This might actually help you with what you are trying to do, according to Jean's answer. In SPSS you need to compute interaction terms for MLR yourself - there is no "> a * b >" button.
Chloe, ANCOVA and regression belong to the same family, they are not so different. You can apply Aiken and West procedure to your data.
Could you give more information about your factors? or send me your data to have a look at it.
Hi Mehmet. Thank you so much for your offer to have a look. Do you have an email address I could contact you on?
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software
- SPSS