Mixed model an linear model gave the same results

#1
I am running some analysis with mixed model with R. I get differents measures from differents persons (person as random effect), during this analysis and looking plots for each people vs measures I did not see too much differences for each person. Then I suspicious that my focus in mixed model is not so good. My model have heterocedastity and residual normality problem. Furthermore looking for variance:

> person = pdLogChol(1)
Variance StdDev
(Intercept) 0.03864167 0.1965749
Residual 3.64527198 1.9092595

I think that variance intercept its so low and residual variance so
high, the person dont give variance in my results??. This is some
results from mi mixed model:


Value Std.Error DF t-value p-value
(Intercept) 0.5962784 0.12821014 2334 4.650789 0.0000
dummy21 -0.8913013 0.24000557 2334 -3.713669 0.0002
countback2 -0.0322950 0.00923287 2334 -3.497829 0.0005
countspace2 0.8046936 0.18837571 2334 4.271748 0.0000
action 0.0001028 0.00001484 2334 6.926781 0.0000
pauseTime 0.0003853 0.00002582 2334 14.923275 0.0000
Duration 0.0007586 0.00003112 2334 24.377110 0.0000
Time 0.0006724 0.00023442 2334 2.868323 0.0042

Also i did run a linear model:


(Intercept) 1.030e+00 1.620e-01 6.356 2.49e-10 ***
person2 -3.257e-01 1.547e-01 -2.106 0.035310 *
person3 -3.560e-01 2.116e-01 -1.683 0.092599 .
person4 -5.931e-01 1.927e-01 -3.078 0.002111 **
person5 -3.800e-01 1.930e-01 -1.969 0.049070 *
person11 -3.261e-01 1.636e-01 -1.993 0.046336 *
person12 -7.539e-01 1.612e-01 -4.678 3.06e-06 ***
person13 -6.464e-01 1.556e-01 -4.155 3.37e-05 ***
person14 -4.125e-01 1.986e-01 -2.077 0.037932 *
dummy21 -8.975e-01 2.402e-01 -3.737 0.000191 ***
countback2 -3.250e-02 9.245e-03 -3.515 0.000448 ***
countspace2 8.031e-01 1.886e-01 4.257 2.15e-05 ***
action 1.030e-04 1.486e-05 6.934 5.29e-12 ***
pauseTime 3.847e-04 2.587e-05 14.868 < 2e-16 ***
Duration 7.588e-04 3.117e-05 24.341 < 2e-16 ***
Time 6.899e-04 2.361e-04 2.922 0.003508 **

From my results I dont see differences for coefficients in both models. Should make my analysis with lm and not focus in mixed models?
Please any comments I grateful.
 

hlsmith

Not a robit
#2
Typically if you think you have a multilevel model, but controlling for groups does not explain any additional variability (ICC), and its not a power issue (sample size), people will revert back to a OLS model and use sandwich SE estimators for simplicity and sandwich for staving off type I errors.