search

modal-inria / MixtComp

Model-based clustering package for mixed data
Other
13 stars 4 forks source link

mean curves out of confidence interval in plotDataCI #2

Closed Quentin62 closed 4 years ago

Quentin62 commented 4 years ago

res.zip

res = readRDS("res.rds")

plotDataCI(res, var = "func", class = 2)

Rplot

Quentin62 commented 4 years ago

the density is defined in the article (section 3.2 formula 7) https://github.com/modal-inria/MixtComp/blob/master/MixtComp/docs/article/functionalModel.pdf

mean is defined by sum_{l=1}^{nSub} kappa_{kl} \sum_{i=1}^{nCoeff} \beta_{kli}* t^{i-1}

for upper and lower bounds, we define the function (say p) implementing formula 7

then, the goal is to find x such that p(x) is equal a requested probability (0.025 and 0.975 for computing a confidence interval). See functionalboundval function

Quentin62 commented 4 years ago

in fact, the optimize function does not found a very close solution.

Instead: Use the uniroot function with an other objective function

Quentin62 commented 4 years ago

correction will be released in RMixtCompUtilities 4.1.3