I am currently using Jamovi to test my hypothesis involving moderator Y on the relationship of A (IV) and B (DV). I am using the linear regression function because there are other variables I need to control, and the medmod function seems to only consider the moderator variable and the independent variable and does account for control variables.
To check my results, I also conducted a linear regression with an interaction term (A x Y) without control variables and compared the results with the results of the moderation analysis in medmod. The results are not far off, but there is a difference (i.e., estimate of interaction term = 0.0103, p = 0.047 using linear regression; estimate = 0.01029, p = 0.043 using medmod).
My question is: what is accounting for the decimal differences? How does Jamovi conduct moderation analysis via medmod vs linear regression with an interaction term?
Linear regression with interaction vs medmod analysis
Re: Linear regression with interaction vs medmod analysis
According to this post:
https://blog.jamovi.org/2017/09/25/medmod.html
medmod uses lavaan to fit a model. I am not sure about the settings it uses for lavaan. If it uses the default options, then maximum likelihood is used as the estimation method. The results can be slightly different from those by Linear Regression, which I believe uses OLS (provided by lm()) to estimate the parameters.
Hope this helps.
-- Shu Fai
https://blog.jamovi.org/2017/09/25/medmod.html
medmod uses lavaan to fit a model. I am not sure about the settings it uses for lavaan. If it uses the default options, then maximum likelihood is used as the estimation method. The results can be slightly different from those by Linear Regression, which I believe uses OLS (provided by lm()) to estimate the parameters.
Hope this helps.
-- Shu Fai