[R-sig-ME] repeated measurements of continuous predictors and continuous response variable

[Slawomir Wierzchon]

Dear all,

I would greatly appreciate receiving exhaustive answer to the next question:

My data consist of one response variable and three predictors, say

Id           y                     x1              x2           x3
1    2,29757646    1,522746484    6,1504    1,81651712
1    2,460906175    2,072253268    7,0467    1,952559422
1    2,561146752    1,69296991    4,8555    1,580112083
2    2,542671567    1,419124506    11,5181    2,443919711
2    2,47840973    1,254903533    10,2239    2,324728117
3    2,409638554    2,168676486    14,7873    2,693768704
4    2,577905082    2,290613721    22,1847    3,099402862
4    2,771908643    1,98419636    18,5161    2,918640624
5    2,137096774    1,166644699    5,1028    1,629789409

Thus there are repeated measurements, and each item (Id) has been subjected
to a diverse number of measurements. I’m interested in finding linear
relationship y = a0 + a1*x1 + a2*x2 + a3*x3.

My questions are:

(a) Can I use complete pooling (i.e. ignore Id’s) to solve my problem? Or
should I use lmer routine? If so, is it sufficient to call it as lmer(y ~
x1+x2+x3 + (1|Id)?

(b)  How can I apply Friedman's “Multivariate Adaptive Regression Splines”
to such problems? That is I'm looking for an R package solving such
problems.

Best regards,

Slawomir

	[[alternative HTML version deleted]]