Hi everyone,
I am running a moderated linear regression and I found a significant interaction effect.
I then used the Preacher online calculator to conduct a simple slopes analysis.
My simple slopes analysis showed that the slope of the regression line at +1SD, mean, -1SD was significantly different from zero (they all show a positive relationship between predictor and outcome at all three values of the moderator tested).
The values are:
cv1 t=5.02, p < .001
cv2 t = 4.92, p < .001
cv3 t=2.11, p = .035
You can see from the pattern of t values and corresponding p values that (as my research team hypothesized), the strength of the relationship between predictor and outcome is stronger at lower levels of the moderator (p < .001 at mean and -1SD compared to p = .035 at +1SD).
However, given that the simple slopes analysis only tells us that the LINE is significantly different from ZERO (and all three are at p < .05)... I am wondering how to compare these lines to each other.
Is there a way to test the lines to illustrate that the slope is steeper for lower levels of my moderator (compared to higher levels of my moderator)? This is what I think is happening in my data -- but I am not sure I can conclude it only from the the different p values in the simple slopes analysis (p = .035 vs. p < .001).
Thanks in advance for the help or advice!
Erin
Get a decent regression program and get confidence intervals on the slopes. see attached for free software and instructions. Best, David Booth
thanks for the info re R! I would definitely like to learn it at some point. But does anyone know if there a way to do this in SPSS or perhaps the online tool?
Hi Erin, you can use modified analisys of variance for compare linear regression. Look my papers, reference.
Interaction terms in typical regression coding schemes (e.g., dummy coding [which is the default in most software] or deviation coding) provide precisely what you are asking for.
AS Stephen suggests see this attachment and my original one to solve the problem. You have a FACTORIAL design
Best wishes, David Booth
Look at https://www.researchgate.net/post/How_to_conduct_simple_slope_for_moderated_regression_analysis
You may by can't obsereved significant effect between two adjacent levels of moderated variable. but... still the moderated effect is kept on.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Regression