|
|
 |
| time scores for time unstructured data |
 
|
| Message/Author |
|
|
| Anh Hua posted on Wednesday, July 08, 2020 - 7:53 am
|
|
|
Dear Drs. Muthen,
I want to conduct a 3 piece multilevel growth model (each student has 9 data points with 3 data points in each year so 1 slope for each year). But within each data wave, students were tested on different dates. Here's an example of my data:
ID data_wave months_of_instruction score
1 1 0.13 214
1 2 4.47 215
1 3 8.28 213
1 4 10.95 214
1 5 14.17 217
1 6 17.62 230
1 7 19.43 230
1 8 23.11 226
1 9 27.32 235
2 1 0.76 197
2 2 4.44 199
2 3 8.97 215
2 4 10.03 207
2 5 14.10 216
2 6 18.44 228
2 7 20.25 216
2 8 24.43 214
2 9 28.83 223
Q1. I first conducted piecewise using data wave as my time variable but I think this assumes that everyone was tested at the same time for all occasions, right, since everyone had the same numeric value for data_wave?
Q2. If I'm right in that I should use months of instruction as my time variable, which ranges approximately from 0 months to 27 months, what woudl my time scores look like?
Here's my time scores for when dave_wave was used as time:
DEFINE:
slope13=0; slope36=0;
IF data_wave GE 3 THEN slope13=2;
IF data_wave GE 3 THEN slope36=data_wave-3;
IF data_wave LE 2 THEN slope13=data_wave-1;
IF data_wave LE 2 THEN slope36=0;
Thank you so much for any help you can provide.
Anh |
|
|
|
|
| Take a look at the Tscores option in the UG. |
|
|
| Anh Hua posted on Friday, July 10, 2020 - 6:27 am
|
|
|
Dear Dr. Muthen,
Once again, you came to my rescue! With your advice, I figured it out.
Now, I can move on to my next research questions.
Thank you!
Anh |
|
|
| Anh Hua posted on Saturday, August 01, 2020 - 7:06 am
|
|
|
Dr. Muthen,
Is the approach you showed me on 7.08.2020, i.e., approach with tscores when you have individually varying time assessments, called definition variable approach in the SEM literature?
I'm currently reading "Growth Modeling: Structural Equation and multilevel modeling approaches" by Grimm et al. and the way they talk about the definition variable approach sounds like the method you showed me.
Also, I noticed that when using tscores, my output does not have indices of global fit such as Chi square, etc but only loglikelihood, AIC, BIC, and sample size adjusted BIC. My understanding is that this is a limitation of the definition variable approach. Am I correct?
Thank you for any clarification you can provide. |
|
|
|
|
| I am not familiar with that term used with Tscores - you may want to ask Grimm. Yes, with Tscores you are essentially doing two-level growth modeling with a random slope for which you don't get fit indices. |
|
|
| Anh Hua posted on Monday, August 03, 2020 - 2:28 am
|
|
|
ahh... thanks so much. I didn't realize that by using TSCORES, I'm essentially doing two-level growth modeling, but this makes sense! This helps me to describe my methododogy in my paper better!
Anh |
|
|
| Back to top |
|
|
