Hello Dr. Geiser, thank you very much for your useful video. I'm wondering what the maximum number of released indicators (e.g., intercepts) is to achieve partial invariance. In other words, how many intercepts per factor can we release?
I'm not sure whether there is an absolute rule. I would guess not since strictly speaking, a single indicator that is invariant would be sufficient (if you knew for sure that this indicator was truly invariant). The problem is that we usually don't know (and we can't test it), so we would have a difficult time trusting in the invariance of a single marker indicator if it is the only indicator that is invariant. For that reason, we typically like to see at least 2 invariant indicators, but again, this is not an absolute rule to my knowledge. Best, Christian Geiser
Hello Dr. Geiser, thank you very much for this video, it was quite useful. I have a query, how can I test the measurement invariance of a scale with categorical items? and specifically, how can the equality constraints of the "thresholds" be imposed or released in case of partial equivalence? A video on this topic would be very helpful. Thank you.
Hello Jorge, Thanks for watching! The principle of imposing and relaxing constraints is the same with categorical (binary and ordinal) indicators. The item threshold parameters are referred to using brackets [] and a $ sign in Mplus. For example, the first threshold for an item u would be referred to as [u$1]; Best, Christian Geiser
![I.N "HALLUCINATION" | [Stray Kids : SKZ-PLAYER]](http://i.ytimg.com/vi/n5B5q1Hwt_U/mqdefault.jpg)
Hello Dr. Geiser, thank you very much for your useful video. I'm wondering what the maximum number of released indicators (e.g., intercepts) is to achieve partial invariance. In other words, how many intercepts per factor can we release?
I'm not sure whether there is an absolute rule. I would guess not since strictly speaking, a single indicator that is invariant would be sufficient (if you knew for sure that this indicator was truly invariant). The problem is that we usually don't know (and we can't test it), so we would have a difficult time trusting in the invariance of a single marker indicator if it is the only indicator that is invariant. For that reason, we typically like to see at least 2 invariant indicators, but again, this is not an absolute rule to my knowledge.
Best, Christian Geiser
Hello Dr. Geiser, thank you very much for this video, it was quite useful.
I have a query, how can I test the measurement invariance of a scale with categorical items? and specifically, how can the equality constraints of the "thresholds" be imposed or released in case of partial equivalence? A video on this topic would be very helpful.
Thank you.
Hello Jorge, Thanks for watching!
The principle of imposing and relaxing constraints is the same with categorical (binary and ordinal) indicators. The item threshold parameters are referred to using brackets [] and a $ sign in Mplus. For example, the first threshold for an item u would be referred to as
[u$1];
Best, Christian Geiser
@@QuantFish Thanks!