Check the phenotype distribution
The phenotype appears to be skewed, so it will need to be corrected. CLARITE makes it easy to plot distributions and to transform variables.
# Discovery
# Make plot to visulize phneotype not log transformed vs. log transfomred.
hist(discovery_cont_final$BMXBMI, breaks = 100)
skewness(discovery_cont_final$BMXBMI)
#> [1] 1.131507
transBMI <- log(discovery_cont_final$BMXBMI)
hist(transBMI, breaks = 100)
skewness(transBMI)
#> [1] 0.390092
# Replace BMI with log BMI
discovery_cont_final$logBMI <- transBMI
discovery_cont_final <- colfilter(discovery_cont_final, cols=c("BMXBMI"),exclude = TRUE)
#> [1] "Running..."
#> [1] "Finished in 0 secs"
# Replication
# Make plot to visulize phneotype not log transformed vs. log transfomred.
hist(replication_cont_final$BMXBMI, breaks = 100)
skewness(replication_cont_final$BMXBMI)
#> [1] 1.522081
transBMI <- log(replication_cont_final$BMXBMI)
hist(transBMI, breaks = 100)
skewness(transBMI)
#> [1] 0.4349248
# Replace BMI with log BMI
replication_cont_final$logBMI <- transBMI
replication_cont_final <- colfilter(replication_cont_final, cols=c("BMXBMI"),exclude = TRUE)
#> [1] "Running..."
#> [1] "Finished in 0 secs"
# Update phenotype name
phenotype <- "logBMI"Create final dataframes
All variables and survey information are combined into one dataframe
discovery_final <- merge_data(discovery_catbin_final, discovery_cont_final, union = FALSE)
#> [1] "Running..."
#> [1] "Finished in 0.023936 secs"
replication_final <- merge_data(replication_catbin_final, replication_cont_final, union = FALSE)
#> [1] "Running..."
#> [1] "Finished in 0.025931 secs"
discovery_final <- merge_data(discovery_final, survey_design_discovery, union=FALSE)
#> [1] "Running..."
#> [1] "Finished in 0.027925 secs"
replication_final <- merge_data(replication_final, survey_design_replication, union=FALSE)
#> [1] "Running..."
#> [1] "Finished in 0.02992 secs"Filter out variables that don’t have a weight both discovery and replication
# Check for weights
catbin_has_weight <- sapply(discovery_catbin_vars,
function(v){ifelse(is.null(weights_discovery[[v]]) | is.null(weights_replication[[v]]),
FALSE,
(weights_discovery[[v]] %in% names(discovery_final)) & (weights_replication[[v]] %in% names(replication_final)))}
)
cont_has_weight <- sapply(discovery_cont_vars,
function(v){ifelse(is.null(weights_discovery[[v]]) | is.null(weights_replication[[v]]),
FALSE,
(weights_discovery[[v]] %in% names(discovery_final)) & (weights_replication[[v]] %in% names(replication_final)))}
)
# Log
if(sum(catbin_has_weight) < length(catbin_has_weight)){
print(paste("Skipped", sum(!catbin_has_weight), "cat/bin variable(s) with no weight: ",
paste(discovery_catbin_vars[!catbin_has_weight], collapse=", ")))}
if(sum(cont_has_weight) < length(cont_has_weight)){
print(paste("Skipped", sum(!cont_has_weight), "continuous variable(s) with no weight: ",
paste(discovery_cont_vars[!cont_has_weight], collapse=", ")))}
#> [1] "Skipped 21 continuous variable(s) with no weight: LBXPFOA, LBXPFOS, LBXPFHS, LBXEPAH, LBXMPAH, LBXPFDE, LBXPFHP, LBXPFNA, LBXPFSA, LBXPFUA, URXP08, URX14D, URX1TB, URX3TB, URXPCP, URXOPP, URXUIO, URXP17, URXP19, DR1TMOIS, LBXPFDO"
# Update
# NOTE: This does it for replication also, since they are copied from these vectors
discovery_catbin_vars <- discovery_catbin_vars[catbin_has_weight]
discovery_cont_vars <- discovery_cont_vars[cont_has_weight]