Message/Author |
|
Brett Foley posted on Wednesday, December 01, 2004 - 8:11 am
|
|
|
Is it possible to fit logistic (“S-shaped”) growth trajectories in Mplus? |
|
bmuthen posted on Wednesday, December 01, 2004 - 8:31 am
|
|
|
Yes, if you give the fixed time scores for such a logistic curve shape. The model is then still linear in the growth factors. No, if you want to estimate a logistic curve where the growth factors (the random effects) enter non-linearly as is sometimes done in the literature. |
|
Brett Foley posted on Wednesday, December 01, 2004 - 11:30 am
|
|
|
Thanks for your quick response. |
|
|
I'm performing LCGA on symptom measures from psychological treatment data. I'm determining the shape of the growth curve before I start investigating the number of classes that best fit the data. As we expect a log-linear relationship between session number and improvement, I have taken the log to base 0.5 (as my measure of treatment response is symptom severity, not improvement) for each timepoint (1 to 12) and used these to specify the gradient as follows: MODEL: int slope | phq_t1@0 phq_t2@-1.00 phq_t3@-1.58 phq_t4@-2 phq_t5@-2.32 phq_t6@-2.58 phq_t7@-2.8 phq_t8@-3.00 phq_t9@-3.17 phq_t10@-3.32 phq_t11@-3.46 phq_t12@-3.58; int WITH slope; phq_t1 - phq_t11 PWITH phq_t2 - phq_t12; Is this correct? Or should I also specify a linear component and a correlation with this, like we do for quadratic growth curves? I don't really understand why this is done. e.g. int lin slope | phq_t1@0... int WITH slope; lin WITH slope; Many thanks! |
|
|
It is not necessary to add to your int slope specification because it covers the development you are interested in. |
|
|
Great, many thanks! |
|
Back to top |