Thanks for your answers. I have been able to obtain individual-time profiles following your advice. I have essentially used:
Below I summarise my experience and ask some questions.
1) I used AR1 model, because my time variable is discrete. I am worried about overfitting, so I payed close attention to precision estimates. I did not notice major changes in precision for the error term after adding time component into the model. However, precision for the time component tends to be very large. Rho for AR1 is very close to 1. When I checked use of RW1 or RW2 I obtained crazy high precision estimates for the time component. Do these observations suggest overfitting?
f(IndI, model="iid")
mean sd 0.025q 0.5q 0.975q mode
Precision for the Gaussian observations 40.51 1.571 37.479 40.49 43.67 40.46
Precision for IndI 46.25 8.810 31.089 45.54 65.62 44.21
f(IndI, model="iid") + f(TimeI, model="iid")
mean sd 0.025q 0.5q 0.975q mode
Precision for the Gaussian observations 42.48 1.739 39.142 42.46 45.99 42.42
Precision for IndI 47.45 9.032 31.908 46.72 67.31 45.35
Precision for DateYearMonthI 925.87 495.161 369.498 796.28 2221.26 613.05
f(IndI, model="iid") + f(TimeI, model="ar1")
mean sd 0.025q 0.5q 0.975q mode
Precision for the Gaussian observations 41.21 1.6061 38.1346 41.190 44.457 41.14
Precision for IndI 46.60 8.8544 31.3679 45.891 66.078 44.55
Precision for TimeI 2538.43 2325.1339 388.2139 1874.091 8696.255 1000.28
Rho for TimeI 0.97 0.0308 0.8879 0.980 0.997 0.99
f(IndI, model="iid") + f(TimeI, model="rw1")
mean sd 0.025q 0.5q 0.975q mode
Precision for the Gaussian observations 41.25 1.610 38.21 41.20 44.52 41.10
Precision for IndI 46.38 8.864 31.387 45.56 66.13 43.96
Precision for DateYearMonthI 37771.18 21158.603 10537.16 33394.37 90959.30 24934.59
f(IndI, model="iid") + f(TimeI, model="rw2")
mean sd 0.025q 0.5q 0.975q mode
Precision for the Gaussian observations 41.22 1.606 38.115 41.20 44.44 41.18
Precision for IndI 46.51 8.826 31.037 45.93 65.60 44.90
Precision for DateYearMonthI 109054.58 42294.622 45722.114 102877.11 209546.46 90408.61
f(IndI, model="iid") + f(TimeI, model="ar1", replicate=IndI)
mean sd 0.025q 0.5q 0.975q mode
Precision for the Gaussian observations 42.81 1.7652 39.4432 42.776 46.3877 42.7084
Precision for IndI 54.75 13.5578 33.4403 52.927 86.3869 49.3999
Precision for TimeI 299.02 195.4974 65.5443 253.903 802.9862 168.2894
Rho for TimeI 0.98 0.0113 0.9557 0.988 0.9976 0.9933
f(IndI, model="iid") + f(TimeI, model="rw1", replicate=IndI)
mean sd 0.025q 0.5q 0.975q mode
Precision for the Gaussian observations 42.96 1.749 39.613 42.92 46.49 42.86
Precision for IndI 50.88 10.404 33.391 49.88 74.18 47.95
Precision for TimeI 13800.06 5888.249 6042.377 12564.92 28610.86 10532.02
f(IndI, model="iid") + f(TimeI, model="rw2", replicate=IndI)
mean sd 0.025q 0.5q 0.975q mode
Precision for the Gaussian observations 43.33 1.748 4.001e+01 43.30 46.88 43.22
Precision for IndI 91.79 31.919 4.393e+01 86.96 167.97 77.95
Precision for TimeI 462998.47 78008.863 3.252e+05 458146.23 630159.17 449701.81
2) Is my understanding correct that with f(x, replicate=y) the order of results in summary.random$x is
...
3) I noticed that when I had Ind and Time variables formatted as factors (I defined Time as either Year-Month, Year-WeekOfTheYear, or Year-DayOfTheYear, with appropriate set of levels, i.e., levels covered all possible Year-SubYear combinations, I provide function to construct such factors below) I got this error: