diff --git a/R/rsimilarity.R b/R/rsimilarity.R index 2a588f98ea76d6aef9c61900c28503c3622c638f..5553a479ee16716f672fe28a1744700593879ad8 100644 --- a/R/rsimilarity.R +++ b/R/rsimilarity.R @@ -53,7 +53,7 @@ #' @importFrom stats lm predict complete.cases cor as.formula #' @export rsimilarity <- function(Rn, predictors, newpredictors, FUN=invRMSE, power=0.5, symmetrize=mean, model="glmnet", - args_glmnet=list(alpha=0.5, s="lambda.min", lower.limits=0), verbose = TRUE, seed=NULL){ + args_glmnet=list(s="lambda.min", lower.limits=0), verbose = TRUE, seed=NULL){ # Checks inputs if(inherits(Rn, "units")) Rn <- units::drop_units(Rn) @@ -112,7 +112,7 @@ rsimilarity <- function(Rn, predictors, newpredictors, FUN=invRMSE, power=0.5, s if(is.character(args_glmnet[["s"]])){s <- glm_similarity[[args_glmnet[["s"]]]]}else{s <- args_glmnet[["s"]]} similarity <- as.vector(stats::predict(glm_similarity, newx=as.matrix(newpredictors), s=s)) coef <- as.matrix(coef(glm_similarity, s=s))[-1,] - # Infinite similarity of predictors leads to infinite predicted similarity (instead of NA) + # Infinite similarity of predictors leads to infinite predicted similarity (instead of NA) inf_similarity <- apply(newpredictors, MARGIN=1, FUN = function(x) any(is.infinite(x))) if(any(inf_similarity)) similarity[inf_similarity] <- Inf # Compute R2 manually @@ -139,7 +139,7 @@ rsimilarity <- function(Rn, predictors, newpredictors, FUN=invRMSE, power=0.5, s coef <- as.matrix(coef(lm_similarity))[-1,] r2 <- summary(lm_similarity)$r.squared pvalue <- summary(lm_similarity)$coefficients[-1, "Pr(>|t|)"] - # Infinite similarity of predictors leads to infinite predicted similarity (instead of NA) + # Infinite similarity of predictors leads to infinite predicted similarity (instead of NA) inf_similarity <- apply(newpredictors, MARGIN=1, FUN = function(x) any(is.infinite(x))) if(any(inf_similarity)) similarity[inf_similarity] <- Inf # Indicators of the distribution of the weights