SamCAndrew
commented
6 years ago I wrote a function to convert austraits into a matrix with each species as a row and each column as a trait. I also divided the traits into types so max and min values for traits have separate columns (this can be easily changed by editing the "value_type"column to be the same for all values for a trait). for numeric traits I took the mean of the values for each species with data. For categorical traits I took the first trait value for each species. It looks like for most traits there are many species with multiple records reporting different trait values. Below is a table summarizing this problem (zero is for species with no data and one means there is only one value reported for these species etc.).
What can we do about this?
The function outputs a list with four data frames. The first data frame is the trait values for each species The second data frame is the number of data points for each species The third data frame is the number of values per species for categorical traits The fourth data frame is the SD values for the means from numeric traits
below is the code for the function.
#### function for Long to wide Format using austraits ####
## set up data
#rm(list = ls())
library(plyr)
library(tidyr)
austraits <- remake::make('austraits')
austraits_data <- austraits$data
list_traits <- unique(austraits_data$trait_name)
## the Function
long_to_wide <- function(list_traits, austraits_data) {
### Prep data
austraits_data[austraits_data == "NA"] <- NA
austraits_data1 <- austraits_data[!is.na(austraits_data$value),]
austraits_data1 <- austraits_data1[austraits_data1$trait_name %in% list_traits, ]
### make trait names
austraits_data1$trait_type <- paste(austraits_data1$trait_name, austraits_data1$value_type, sep = "_")
list_traits2 <- as.data.frame(unique(paste(austraits_data1$trait_name, austraits_data1$value_type, sep = "/")))
continuous_traits <- c("leaf_area", "leaf_length", "leaf_width", "leaf_area_per_sapwood_area", "leaf_area_ratio",
"seed_mass", "seed_volume", "seed_length", "seed_width", "wood_density",
"seed_breadth", "leaf_C_per_dry_mass", "leaf_cell_wall_fraction", "leaf_cell_wall_N",
"bark_mass_area", "bark_thickness", "branch_mass_fraction", "diaspore_mass", "flowering_time",
"fruit_diameter", "fruit_length", "fruiting_time", "genome_size", "glaucous",
"leaf_cell_wall_N_fraction", "leaf_chlorophyll_per_dry_mass", "leaf_CN_ratio", "leaf_dark_respiration_per_area", "leaf_dark_respiration_per_dry_mass",
"leaf_delta13C", "leaf_delta15N", "leaf_dry_mass", "leaf_dry_matter_content", "leaf_elastic_modulus",
"leaf_fracture_force", "leaf_K_per_area", "leaf_K_per_dry_mass", "lignotuber", "leaf_lifespan",
"leaf_lobed", "leaf_mass_fraction", "leaf_N_per_area", "leaf_N_per_dry_mass", "leaf_P_per_area",
"leaf_P_per_dry_mass", "leaf_photosynthetic_nitrogen_use_efficiency", "leaf_photosynthetic_water_use_efficiency", "leaf_saturated_water_content_per_mass", "leaf_specific_conductivity",
"leaf_thickness", "leaf_toughness", "leaf_water_content_per_area", "leaf_water_content_per_mass", "nitrogen_fixing",
"photosynthetic_rate_per_area", "photosynthetic_rate_per_dry_mass", "plant_height", "root_wood_density", "sapwood_specific_conductivity",
"seed_mass_reserve", "seed_P_concentration", "seed_size_less_than_10mm", "seed_size_over_10mm", "serotiny",
"specific_leaf_area", "specific_leaf_area_SE", "stem_cross_section_area", "stem_cross_section_area_no_bark", "stem_density",
"stem_tissue_density", "stomatal_conductance_per_area", "storage_organ_present", "succulent", "water_use_efficiency")
Categorical_traits <- c("aquatic", "dispersal_syndrome", "fire_response", "flower_colour", "fruit_type",
"fruit_type_botany", "fruit_type_function", "growth_habit", "leaf_arrangement", "leaf_compoundness",
"leaf_margin", "leaf_phenology", "leaf_shape", "leaf_type", "life_history",
"lifespan", "lineage", "photosynthetic_pathway", "plant_growth_form", "regen_strategy",
"root_structure", "seed_shape", "woodiness", "dispersal_appendage")
Species_list <- as.data.frame(unique(austraits_data1$species_name))
colnames(Species_list) <- "sp_name"
aus_continuous <- austraits_data1[austraits_data1$trait_name %in% continuous_traits,]
aus_continuous$value1 <- as.numeric(aus_continuous$value)
#### same thing for all birds just without "sex_M.1" as a extra subsetting factor
Trait_by_species<-ddply(aus_continuous,c("trait_type", "species_name"),summarise,
N_pop=sum(!is.na(value1)), trait_mean=mean(value1),
SD_trait=sd(value1))
### Get missing data
missing_Species_1 <- as.data.frame(Species_list[!Species_list$sp_name %in% Trait_by_species$species_name,])
colnames(missing_Species_1) <- "species_name"
matrix_1 <- as.data.frame(matrix( ,nrow(missing_Species_1), length(unique(Trait_by_species$trait_type))))
colnames(matrix_1) <- c(unique(Trait_by_species$trait_type))
missing_Species_1 <- cbind(missing_Species_1, matrix_1)
## mean
Trait_mean <- Trait_by_species[,c("trait_type", "species_name", "trait_mean")]
mean_wide <- spread(Trait_mean, trait_type, trait_mean)
mean_wide <- rbind(mean_wide, missing_Species_1)
mean_wide <- mean_wide[order(mean_wide$species_name),]
## n
Trait_n <- Trait_by_species[,c("trait_type", "species_name", "N_pop")]
n_wide <- spread(Trait_n, trait_type, N_pop)
n_wide <- rbind(n_wide, missing_Species_1)
n_wide <- n_wide[order(n_wide$species_name),]
## SD
Trait_SD <- Trait_by_species[,c("trait_type", "species_name", "SD_trait")]
SD_wide <- spread(Trait_SD, trait_type, SD_trait)
SD_wide <- rbind(SD_wide, missing_Species_1)
SD_wide <- n_wide[order(SD_wide$species_name),]
SD_for_values <- SD_wide
rm(Trait_mean, Trait_n, Trait_SD, missing_Species_1, matrix_1)
aus_categorical <- austraits_data1[austraits_data1$trait_name %in% Categorical_traits,]
Trait_by_species2<-ddply(aus_categorical,c("trait_name", "species_name"),summarise,
N_pop=sum(!is.na(value)))#, trait_mean=paste0(unique(value)))
missing_Species_2 <- as.data.frame(Species_list[!Species_list$sp_name %in% Trait_by_species2$species_name,])
colnames(missing_Species_2) <- "species_name"
matrix_2 <- as.data.frame(matrix( ,nrow(missing_Species_2), length(unique(Trait_by_species2$trait_name))))
colnames(matrix_2) <- c(unique(Trait_by_species2$trait_name))
missing_Species_2 <- cbind(missing_Species_2, matrix_2)
## n
n_wide_2 <- spread(Trait_by_species2, trait_name, N_pop)
n_wide_2 <- rbind(n_wide_2, missing_Species_2)
n_wide_2 <- n_wide_2[order(n_wide_2$species_name),]
Species_list2 <- as.character(Species_list$sp_name)
Cat_traits<- c(unique(aus_categorical$trait_name))
all_traits <- list()
for(i in 1:length(Cat_traits)){
trait <- aus_categorical[aus_categorical$trait_name == Cat_traits[i],]
df <- list()
for(ii in 1:length(Species_list2)){
species <- trait[trait$species_name == Species_list2[ii],]
sp <- as.data.frame(Cat_traits[i])
colnames(sp) <- "trait_name"
sp$species_name <- Species_list2[ii]
sp$trait_value <- if(nrow(species)>0) { paste0(unique(species$value)[1]) } else {NA}
sp$no_trait_values <- length(unique(species$value))
df[[ii]] <- sp
}
all_traits[[i]] <- do.call("rbind",df)
}
rm(i, ii, list_traits, Categorical_traits, continuous_traits)
Trait_mean2 <- do.call("rbind", all_traits)
Trait_mean3 <- Trait_mean2[,c("trait_name", "species_name", "trait_value")]
mean_wide_2 <- spread(Trait_mean3, trait_name, trait_value)
mean_wide_2 <- mean_wide_2[order(mean_wide_2$species_name),]
Trait_mean4 <- Trait_mean2[,c("trait_name", "species_name", "no_trait_values")]
mean_n_trait_2 <- spread(Trait_mean4, trait_name, no_trait_values)
mean_n_trait_2 <- mean_n_trait_2[order(mean_n_trait_2$species_name),]
rm(Species_list2, all_traits, missing_Species_2, matrix_2, trait, df, species, sp,
Trait_mean3, Trait_mean4)
trait_values <- cbind(mean_wide, mean_wide_2[,2:length(mean_wide_2)])
number_of_observations <- cbind(n_wide, n_wide_2[,2:length(n_wide_2)])
number_of_catergorical_values <- mean_n_trait_2
rm(mean_wide, mean_wide_2, n_wide, n_wide_2, mean_n_trait_2, SD_wide,
Trait_by_species, Trait_by_species2, Trait_mean2, Species_list)
output <- list()
output[[1]] <- trait_values ## trait values for each species
output[[2]] <- number_of_observations ## number of data points for each species
output[[3]] <- number_of_catergorical_values ## number of values per species for categorical traits
output[[4]] <- SD_for_values ## SD for means calculated for numeric traits
return(output)
}
dfalster
(and give correct value types: character, integer, etc)