I would like to compare whether the slope is statistically different between two groups while pretending that the intercept was exactly the same between the groups.
A bit about the model: I run a growth curve model with economic resources (continuous variables in Euro) as my outcome measure. My explanatory variables are time since divorce ("divduration" in years) and whether a respondent actually experienced a divorce ("treat" with 1=divorced, 0=continuously married). So the base model looks as follows:
Code:
mi estimate, dots post: mixed wealth c.divduration##i.treat $control1 || id: divduration if psmatched2 ==1, variance mle
Code:
mi estimate, dots post: mixed wealth c.divduration##i.treat##i.remar $control1 || id: divduration if psmatched2 ==1, variance mle
Code:
Imputations (5): ..... done Multiple-imputation estimates Imputations = 5 Mixed-effects ML regression Number of obs = 9,760 Group variable: id Number of groups = 5,006 Obs per group: min = 1 avg = 1.9 max = 4 Average RVI = 6.4175 Largest FMI = 0.9839 DF adjustment: Large sample DF: min = 4.18 avg = 64.83 max = 397.52 Model F test: Equal FMI F( 7, 134.7) = 11.17 Prob > F = 0.0000 ------------------------------------------------------------------------------------------- wealth | Coef. Std. Err. t P>|t| [95% Conf. Interval] --------------------------+---------------------------------------------------------------- divduration | 2221.492 643.0812 3.45 0.003 874.101 3568.884 | treat | Treated | -39046.84 8871.472 -4.40 0.000 -56692.11 -21401.57 | treat#c.divduration | Treated | -297.1171 1363.105 -0.22 0.828 -2976.913 2382.679 | 1.remar | 9413.24 16072.45 0.59 0.562 -23172.01 41998.5 | remar#c.divduration | 1 | 1394.398 2100.509 0.66 0.509 -2796.951 5585.746 | treat#remar | Control#1 | 0 (empty) Treated#1 | 0 (omitted) | treat#remar#c.divduration | Control#1 | 0 (empty) Treated#1 | 0 (omitted) | 1.flag_firstwealth | -13878.67 4438.822 -3.13 0.003 -22757.7 -4999.634 1.flag_impwealth | 14860.63 6932.212 2.14 0.067 -1366.448 31087.72 _cons | 81025.73 5520.197 14.68 0.000 69689.74 92361.72 ------------------------------------------------------------------------------------------- ------------------------------------------------------------------------------ Random-effects Parameters | Estimate Std. Err. [95% Conf. Interval] -----------------------------+------------------------------------------------ id: Independent | sd(divduration) | 11011.74 1056.377 8637.166 14039.15 sd(_cons) | 99006.4 6453.131 84522.53 115972.3 -----------------------------+------------------------------------------------ sd(Residual) | 133315.1 9924.388 108806.7 163344 ------------------------------------------------------------------------------
My problem/my question: I would now like to test whether there is a statistical difference between the growth curve of divorcees that are ever remarried compared to divorcees that are never-remarried while keeping their initial differences constant. So at the moment, remarried divorcees have 1394 Euros more right at divorce than never-married divorcees. I would like to pretend that there was no initial difference and then test whether the two differ in their growth rate after divorce.
Note: continuously married respondents can never be remarried and thus their interaction coefficient falls out of the model.
Data: SOEP
STATA: 16.1
I am not very experienced with postestimation commands and appreciate any advice on how I can solve this.
Thank you,
Nicole
0 Response to Comparing coefficients while keeping difference between other coefficients constant in mixed model
Post a Comment