|
|
 |
| Latent Trait Development in Panel Data |
 
|
| Message/Author |
|
|
| Stefan P. posted on Tuesday, June 30, 2020 - 8:10 am
|
|
|
Dear Dres. M&M,
I would like to determine the influence of age on the development of the five personality traits per SEM. I have panel data, the big five are measured at four points in time every four years with a total of 15 items. Specht et. al. (2011) have performed a growth curve analysis in their paper "Stability and change of personality across the life course: the impact of age and major life events on mean-level and rank-order". But I wonder whether this is really the right approach. After all, I do not primarily want to know how traits change over the course of 12 years, but rather as a function of age - although repeated measures and individuals nested in households should be taken into account. Is this really a case for LGCM or should one rather work with clusters?
Is there a suitable trainig video on the website?
Many thanks and best regards
Stefan |
|
|
|
|
| Using age instead of occasion is shown in UG ex 6.18. It is discussed in our Short Course Topics 3 and 4. |
|
|
| Stefan P. posted on Thursday, July 02, 2020 - 5:08 am
|
|
|
Thank very much.
hmmmm - I was afraid of that. That means, if I have 4 observations and in each observation, i have birth years from about 1920 to 2000, I must form four groups (for the observations) ...
In addition I have to enter the assignments age<-->observation in a matrix first, so that I can manually define about 80 years (age) in each group.
And this I have to do for each of the five personality traits in a separate model ... right?
Is there any place in the UG where the DATA COHORT command is explained in more detail (except for pp. 592)?
Thanks for the effort
Stefan |
|
|
| Stefan P. posted on Thursday, July 02, 2020 - 1:28 pm
|
|
|
sorry - I'm afraid I mixed up groups and observations...
I think I need 91 groups (birth year 1909-1999) and then I have 4 observations in each group (Survey years 2005, 2009, 2013, 2017)... right?
Do the variable names a21, a22, a23, a24 in example 6.18 stand for
age in assessment1, age in assessment2, age in assessment3, age in assessment4 or for
time1, time2, time3, time4?
Best regards
Stefan |
|
|
|
|
With such a wide age range, the multi-cohort approach won't work because you don't have much of any age overlap. Maybe just analyze the growth separately for different age categories.
The a variables in 6.18 are not age but time-varying covariates. |
|
|
| Stefan P. posted on Sunday, July 05, 2020 - 1:21 pm
|
|
|
Thank you - but wouldn't it be most practical then to perform a multilevel regression analysis with individual ID as cluster (as briefly outlined in the video for Topic 7: "Multilevel modeling of cross-sectional data"? One could then use age, age^2 and more polnomial terms as covariates (right?)...
Best regards
Stefan |
|
|
|
|
| True, but different growth shapes might be relevant for different age ranges. |
|
|
| Back to top |
|
|
