I have a three-level mixed model and I am struggling on the specific interpretation of the coefficients.
I have observations nested within individuals, further nested within firms. I am trying to understand whether a policy's effectiveness varies with age. The data are in a three-wave panel with two years between each wave. The panel is quite unbalanced with many people appearing only once.
A simplified version of my model and output is below. This is using test data, due to data access rules, so the results may not make a lot of sense. I will also add that I have calculated marginal effects which help me interpret the model practically, but I am unclear precisely how to interpret the regression coefficients with respect to the levels (i.e., within and between person). I have seen other posts on this issue but am still unsure with regard to my own model. I would appreciate any help.
Code:
mixed y c.age##i.policy i.year || firm: || person: age Mixed-effects ML regression Number of obs = 21,398 ------------------------------------------------------------- | No. of Observations per Group Group Variable | Groups Minimum Average Maximum ----------------+-------------------------------------------- firm | 6,502 1 3.3 19 person | 17,582 1 1.2 3 ------------------------------------------------------------- Wald chi2(5) = 3.19 Log likelihood = -23030.701 Prob > chi2 = 0.6707 --------------------------------------------------------------------------------- y | Coef. Std. Err. z P>|z| [95% Conf. Interval] ----------------+---------------------------------------------------------------- age | .0000896 .0015416 0.06 0.954 -.002932 .0031111 | policy | 1 Yes | -.1205814 .0992169 -1.22 0.224 -.3150429 .0738802 | policy#c.age | 1 Yes | .0024477 .0021021 1.16 0.244 -.0016723 .0065676 | year | 2000 | -.0004633 .0119253 -0.04 0.969 -.0238365 .0229099 2002 | -.002491 .0126007 -0.20 0.843 -.027188 .0222059 | _cons | 3.002878 .0716159 41.93 0.000 2.862513 3.143243 --------------------------------------------------------------------------------- ------------------------------------------------------------------------------ Random-effects Parameters | Estimate Std. Err. [95% Conf. Interval] -----------------------------+------------------------------------------------ firm: Identity | sd(_cons) | 1.35e-06 . . . -----------------------------+------------------------------------------------ person: Independent | sd(age) | 6.83e-06 . . . sd(_cons) | .000022 . . . -----------------------------+------------------------------------------------ sd(Residual) | .7098958 . . . ------------------------------------------------------------------------------ LR test vs. linear model: chi2(3) = 0.00 Prob > chi2 = 1.0000 Note: LR test is conservative and provided only for reference.
0 Response to Interpreting coefficients of a three-level mixed model
Post a Comment