HenrikBengtsson/future#130

Link

https://cran.r-project.org/web/packages/fitdistrplus/fitdistrplus.pdf

Purpose

Set the theta prior

How to

Sum of the raw score than test the distribution if calibration works for non-nominal models

Nonparametric Bootstrapping

Traditional researchers believe the randomized samples will get the best result in their research, so that I have to add this ;)

Add start time information in actual results

Citation

Lee, W.-Y., Cho, S.-J., & Sterba, S. K. (2018). Ignoring a Multilevel Structure in Mixture Item Response Models: Impact on Parameter Recovery and Model Selection. Applied Psychological Measurement, 42(2), 136–154. http://doi.org/10.1177/0146621617711999

"Based on these findings, it is concluded that, for class-specific ICC conditions examined here, a multilevel mixture item response model is recommended over a single- level item response model for a clustered dataset having cluster size >20 and the number of clusters >50." (from abstract, p. 136)

Where to fix

https://github.com/seonghobae/kaefa/blob/master/R/utils.R#L59-L75

How to fix

• Add the code if cluster size < 20, exclude the variable from random effect term.

https://stackoverflow.com/questions/9689104/installing-r-on-mac-warning-messages-setting-lc-ctype-failed-using-c

click-by-click interface for applied psychologists

Most applied psychologists are want to easy method to do their research with robust measurement theories. However, they feel hard to use the software. They do not have any experience with making their codes.

http://mapository.tistory.com/entry/%EC%BD%94%EB%93%9C-%EC%BB%A4%EB%B2%84%EB%A6%AC%EC%A7%80

philchalmers/mirt@732a71a

devtools::spell_check(dict = 'en_GB')

pickGlobalOptimal

In current, most aefa related functions are returning the list of possible appropriate solutions. However that’s too hard to select optimal model to general users like applied psychologists. So that I will make a function what calls ‘pickGlobalOptimal’, person can set the criterion indices like the DIC, AIC, BIC, SABIC. As the default, I’ll set the DIC.

seongho@onmyway:~$ cat /etc/os-release | grep NAME
NAME="Ubuntu"
PRETTY_NAME="Ubuntu 16.04.3 LTS"
VERSION_CODENAME=xenial
UBUNTU_CODENAME=xenial

[root@s3 ~]# cat /etc/os-release | grep NAME
NAME="CentOS Linux"
PRETTY_NAME="CentOS Linux 7 (Core)"
CPE_NAME="cpe:/o:centos:centos:7"

Maybe, system('cat /etc/os-release | grep NAME') helps do this.

https://ko.m.wikipedia.org/wiki/탐욕_알고리즘

test that

https://journal.r-project.org/archive/2011-1/RJournal_2011-1_Wickham.pdf

https://github.com/armstrtw/rzmq
http://zguide.zeromq.org/page:all

Details

In current, kaefa:::.covdataFixedEffectComb returns a list which includes the classified categorical variable candidates for fixed and random effect.

So, In currently, Support the auto combination of numeric variables required.

Especially, L9 to L16 was incomplete.

Todo

• Check is that a unique numeric variable which one of the covariance data frame.

• Add return outcomes which one is real numeric variable (ex. age). -- If False Positive, That doesn't matter in estimation. Because that's just a trial, not harm.

• Link with combination function in internal.
  

  kaefa/R/newEngine.R

  Lines 75 to 85 in da6bf9d

  	combine <- function(x, y) {
  	combn(y, x, paste, collapse = ", ")
  	}
  	if (ranefautocomb) {
  	randomEffectCandidates <- paste0("list(", unlist(lapply(0:NROW(random), combine,
  	random)), ")")
  	} else {
  	randomEffectCandidates <- paste0("list(", unlist(lapply(NROW(random):NROW(random),
  	combine, random)), ")")
  	}

  
  

  kaefa/R/kaefa.R

  Line 383 in da6bf9d

  BURNIN = 1500, SEMCYCLES = 1000, covdata = NULL, fixed = kaefa:::.covdataFixedEffectComb(covdata), random = lapply(c(kaefa:::.covdataClassifieder(covdata)$random, 'items'), FUN = function(X){eval(parse(text = paste0('as.formula(',paste0('~1|',X), ')')))}), key = NULL, accelerate = "squarem", symmetric = F, saveModelHistory = T,