StudCIaddition

[IlmariTamminen]

Here is the first part of the studentized-CI implementation. The bootstrap.merMod has an option to acquire the needed variances. Try using varest argument with “no” (default), “analyticalf” (obtains variances from vcov matrix), or “nestboot” (not implemented yet, only placeholder). The “analyticalf” works with .f() returning named nums or dataframes. The variances appear in the replicates of the result. The calculation of the confidence intervals themselves are not included yet.

[aloy]

Sorry that this took me so long to get to, summer has been far busier than anticipated. It looks like you are thinking about this approach only for the standard coefficient and variance component estimates. Is that correct?

[IlmariTamminen]

Hi @aloy,

No problem, I have also been busy with other tasks.

Yes, with extra place holder for nested bootsrapping for any other statistic as well.

To my understanding, the calculation of the studentized CIs require the statistic of interest, its variance, and the distributions of the replicates and their variances. The argument "analyticalf" can be used with the basic .f = fixef. I planned that when later the confint() is called, the user defines which fixed effect and variance (and the replication distributions) are used for the studentized CI.

As an extra feature, I prepared the option also for the nested bootstrapping ("nestboot") that can be used to estimate the variance for any custom statistics, not just for fixef.