[R] Comparison of two penalized spline fits in mixed model framework

[Anna-Leena Orsama]

Hello!

I have faced a problem in nlme-environment. My intention is to fit a  
penalized spline  model in mixed model framework. I want make a  
comparison in smooth curves between two groups but for some reason I  
get NaN in output..

Hereis the R-code I have used.

#Z.overall is for truncated lines part
knots <- seq(1.5,6.5,by=1)
Z <- outer(weekday,knots,"-")
Z.overall <- Z*(Z>0)

fit1 <- lme(weight~group*weekday, random=list(group=pdIdent(~Z.overall-1)))
summary(fit1)
Linear mixed-effects model fit by REML
  Data: NULL
        AIC      BIC    logLik
   4379.838 4411.895 -2183.919

Random effects:
  Formula: ~Z.overall - 1 | group
  Structure: Multiple of an Identity
         Z.overall1 Z.overall2 Z.overall3 Z.overall4 Z.overall5  
Z.overall6  Residual
StdDev:  0.1310617  0.1310617  0.1310617  0.1310617  0.1310617   
0.1310617 0.9833188

Fixed effects: normMAweight2 ~ group * weekday
                     Value Std.Error   DF    t-value p-value
(Intercept)    0.23589909 0.4431604 1545  0.5323109  0.5946
group          0.14167744 0.2629714    0  0.5387562     NaN
weekday       -0.08601980 0.2770228 1545 -0.3105152  0.7562
group:weekday -0.02575775 0.1686222 1545 -0.1527542  0.8786
  Correlation:
               (Intr) group  weekdy
group         -0.957
weekday       -0.911  0.883
group:weekday  0.861 -0.915 -0.953

Standardized Within-Group Residuals:
         Min          Q1         Med          Q3         Max
-3.87210364 -0.64218983 -0.03839294  0.60534552  4.30615258

Number of Observations: 1549
Number of Groups: 2
Warning message:
In pt(q, df, lower.tail, log.p) : NaNs produced




Kindly Regards,
Anna-Leena Orsama