Timeline for What to set "intercept factor" and "slope factor" as in model?
Current License: CC BY-SA 4.0
6 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Oct 19, 2023 at 12:54 | comment | added | Matt W | Thanks Christian, this is really helpful. | |
| Oct 19, 2023 at 0:42 | comment | added | Christian Geiser | Also, you have to make sure your simulated values for the latent factor variances and the error variances are plausible/realistic. This can be more tricky, especially with regard to the slope factor variance. | |
| Oct 19, 2023 at 0:41 | comment | added | Christian Geiser | The means of the intercept and slope refer to the means of the latent factors. They are not identical to the observed variable means. However, the observed means can serve as a guide for finding plausible/realistic values for the intercept and slope factor means. It is plausible that the intercept factor mean is not too far away from the Time 1 observed variable mean. Regarding the slope factor mean, you have to ask yourself whether it is plausible that the average rate of linear decline between two time points is -1. In any case, I would not estimate an LGCM with only N = 10 cases. | |
| Oct 18, 2023 at 16:25 | comment | added | Matt W | So regarding the estimated parameter of the mean intercept and mean slope, does this relate to the observed variable? For example, I expect the average PHQ-9 score in this study to be around 20 at first interval with a moderate variance. However if I change the mean intercept to 20, with an expected slope of -1. The power analysis using the code estimated 10 participants provides 0.88 Power, which just doesn't sound right to me and made me figure my understanding was wrong. Is this actually correct? | |
| Oct 18, 2023 at 11:19 | history | edited | Christian Geiser | CC BY-SA 4.0 |
added 179 characters in body
|
| Oct 18, 2023 at 11:12 | history | answered | Christian Geiser | CC BY-SA 4.0 |