[R] how to specify starting values in varIdent() of lme()

[Douglas Bates]

On 6/1/07, toby909 at gmail.com <toby909 at gmail.com> wrote:
> I was reading the help but just did not get how to specify starting values for
> varIdent() of the lme() function, although I managed to do it for corSymm().

> Do I specify the values just as they are printed out in an output, like c(1,
> 1.3473, 1.0195). Or do I need to take the residual and multiply it with these
> like c(0.2235, 0.2235*1.3473, 0.2235*1.0195)
> or any other form that I dont know of?

Strangely enough you specify it in the form described in

?varIdent

Look at the description of the value argument.

>
> Thanks Toby
>
>
>
>
>
> Linear mixed-effects model fit by REML
>   Data: dtaa
>         AIC       BIC   logLik
>    -788.783 -692.5656 409.3915
>
> Random effects:
>   Formula: ~timef - 1 | orgid
>   Structure: General positive-definite, Log-Cholesky parametrization
>           StdDev     Corr
> timef1   0.04398482 timef1 timef2
> timef2   0.07910354 1
> timef3   0.03648411 1      1
> Residual 0.22350583
>
> Correlation Structure: General
>   Formula: ~1 | orgid/id
>   Parameter estimate(s):
>   Correlation:
>    1     2
> 2 0.487
> 3 0.597 0.440
> Variance function:
>   Structure: Different standard deviations per stratum
>   Formula: ~1 | time
>   Parameter estimates:
>         1        2        3
> 1.000000 1.347341 1.019529
> Fixed effects: score ~ timef - 1
>              Value   Std.Error   DF   t-value p-value
> timef1 -0.3846847 0.007627811 4421 -50.43186       0
> timef2 -0.2727646 0.012012786 4421 -22.70619       0
> timef3 -0.3961244 0.007180147 4421 -55.16939       0
>   Correlation:
>         timef1 timef2
> timef2 0.735
> timef3 0.746  0.670
>
> Standardized Within-Group Residuals:
>         Min         Q1        Med         Q3        Max
> -1.4659548 -0.5982929 -0.4096429  0.3101507  4.0911728
>
> Number of Observations: 4515
> Number of Groups: 92
>
> ______________________________________________
> R-help at stat.math.ethz.ch mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.
>