Hello, thank you so much! I came to know this RU-vid by googling "cross-lagged panel analysis." I was happily surprised to learn a depth and clarity of knowledge you had possessed. Every video is a great help. A single thank-you is absolutely not enough to express my immense gratitude. Look forward to listening to more of your stats knowledge. :) Please stay healthy and safe! Thank you so mmmuch!
This is fascinating. In my field, RCTs are very popular, and most of our stats knowledge are based on this design. I have been confused for a long time about how I understood why distinguishing within and between was relevant when people were talking about it, yet I didn't formally understand why it didn't matter to us. And with your video, it makes sense, in RCTs we find ourselves in a situation where the contextual effect is 0, therefore within effect = between effect.
Glad that you found it useful. I guess if you have a within-person experimental design where each individual is subjected to multiple treatments sequentially, you could have spillover effects so that first treatment affects also the performance (or what ever the dependent variable) after the second treatment. But you generally try to design your experiments to avoid this. And in a well-designed within-person experiment the contextual effect should indeed be 0.
Thank You !!😀 This explanation is very clear and insightful. I have read about within effects and between effects but this really helped improve my understanding.
Hello! Thanks for the explanation, the Woolridge within between estimator (REWB) models generates the between effects as the cluster means and I believe the xthybrid command in STATA does the same thing. However, in your other video you mention that contextual effect is generated by adding cluster means as explanatory variables, so I'm a little confused. Could you please help me understand?
Including clusters means gives either between effect or contextual effect depending on whether you cluster mean center the predictor or not. See page 451 and the references therein in journals.sagepub.com/doi/full/10.1177/1094428119877457
Hi Mikko Rönkkö, would you please suggest me statistcal conditions that one should meet to go for within-between effect modelling. Or, if my question is wrong. How to jsutify if I have used between-within effect model. Can you suggest some reading material for my queries?
That depends on your research question. You can see e.g. table 2 in this paper jyx.jyu.fi/handle/123456789/66704 and then think which of the two effects better corresponds to your research question. The article also cites some literature on multilevel modelling that can be useful.
