random_rpql() allows the user to do variable selection for fixed effect and random effect in GLMMs model by using rpql(Hui,et al,2017) and random lasso(Wang, et al., 2011)
Usage
random_rpql(data,q1_fix,q1_random,bt1,fixed_effect,random_effect,ycol,id_vector,no_scale,cluster_id,within_cluster_id,family="gaussian",lam1,verbose=T,ci_criteria1=5,pen.type1="adl",random_draw_sample1=0.5,q2_fix,q2_random,bt2,lam2,pen.type2="adl",ci_criteria2=5,random_draw_sample2=0.5,use_lmer=F,use_fre_fun1=T)Arguments
- data
dataset without missing data
- q1_fix
number of fixed effects used in each bootstrapping step in first step of random rpql
- q1_random
number of random effects used in each bootstrapping step in first step of random rpql
- bt1
number of bootstrapping in first step of random rpql
- fixed_effect
candidate variables used to fixed effects
- random_effect
candidate variables used to random effects
- ycol
outcome variables
- id_vector
column names for groups (cross-sectional design)/participants id(for longitudinal design) and within-units id
- no_scale
variable names of discrete variables
- cluster_id
variable name for participants id/group id
- within_cluster_id
variable name for time/id for participant within each unit(participant/group)
- family
The distribution for the outcome in GLMM. Supported arguments include: gaussian(), binomial().
- lam1
a vector for tuning parameters used in rPQL for each bootstrapping step in first step
- verbose
show the running details
- ci_criteria1
criteria to choose the optimal tuning parameters for step 1, details refers to Hui,et al,2017
- pen.type1
the type of penalty used in random rpql first step. supported argument include: "lasso" for standard lasso(Tibshirani,1996),"adl"for adaptive lasso(Zou,2006)
- random_draw_sample1
the proportion sample used in each bootstrapping in first step
- q2_fix
number of fixed effects used in each bootstrapping step in second step of random rpql
- q2_random
number of random effects used in each bootstrapping step in second step of random rpql
- bt2
number of bootstrapping in second step of random rpql
- lam2
a vector for tuning parameters used in rPQL for each bootstrapping step in first step
- pen.type2
the type of penalty used in random rpql second step
- ci_criteria2
criteria to choose the optimal tuning parameters for step 2, details refers to Hui,et al,2017
- random_draw_sample2
the proportion sample used in each bootstrapping in second step
- use_lmer
if pen.type2 is true, whether use lmer to derived the weights of penalty. if use_lmer=T, using lmer to calculate weights for "adl". otherwise, the weights for "adl" are calculated by the importance index obtained by first step
- use_fre_fun1
whether used frequency list to calculate importance index
References
Hui, F. K. C., Mueller, S., and Welsh, A.H. (2017). Hierarchical Selection of Fixed and Random Effects in Generalized Linear Mixed Models. Statistica Sinica, 27, 501-518. Wang, S., Nan, B., Rosset, S., & Zhu, J. (2011). Random lasso. The annals of applied statistics, 5(1), 468.
Examples
# \dontrun{
id_vector<- c("participant_id", "wave")
no_scale <- c("participant_id", "wave")
cluster_id="participant_id"
within_cluster_id="wave"
ycol <- "y"
family<-"gaussian"
lam1=exp(seq(from=log(1e-4),to=log(1e-8),length.out=100))
lam2=exp(seq(from=log(1e-1),to=log(1e-8),length.out=100))
random_rpql(data,q1_fix=4,q1_random=3,bt1=200,fixed_effect,random_effect,ycol,id_vector,no_scale,cluster_id,within_cluster_id,family="gaussian",lam1,verbose=T,ci_criteria1=5,pen.type1="adl",random_draw_sample1=0.5,q2_fix=4,q2_random=3,bt2=200,lam2,pen.type2="adl",ci_criteria2=4,random_draw_sample2=0.5,use_lmer=F,use_fre_fun1=T)
#> Error in UseMethod("select"): no applicable method for 'select' applied to an object of class "function"
# }