ks_00_timeseries_traitimpute.Rmd

---
title: "ks_00_traitimpute"
author: "Kate Sheridan"
date: "3/21/2022"
output: html_document
---

This script does imputation from the functional traits csvs with mice.
This differs from previous scripts in that it doesn't make a dissimilarity matrix, cluster, or do any further analysis. It simply removes NAs by imputing traits with their most probable replacement. 

To run the matrix/cluster/etc from the previous script, use ks_03_timeseries_matrix or ks_04_timeseries_cluster-tsne.
But our current analysis uses only MCA/HCPC

```{r setup, include=FALSE}
library(readr)
library(dplyr)
library(mice)
library(here)
```

```{r loadin}
#be sure to use na.strings or imputation will fail!

# algae traits not ready
algae_raw <- read.csv(here("data", '20220525_shoals_algae_traits.csv'), 
                      na.strings = c("","NA", 'N/A')) %>% filter(!(is.na(species)))

animal_raw <- read.csv(here("data", '20220320_shoals_animal_traits.csv'), 
                      na.strings = c("","NA", 'N/A')) %>% filter(!(is.na(species)))
```

# Select columns 

Only columns that will be used later; remove notes, references, etc.
Also remove any columns known a priori to be collinear, or move them to the front to be masked during imputations.


```{r filter}
algae_raw <- algae_raw %>%
  select(species, phylum, group, common_name_division_bgr, invasive_gom,
         intertidal_gom, subtidal_gom, autotroph,
         # s-d morphology, expanded
         erect_morphology_growth_form, foliose_morphology_growth_form,
         filamentous_morphology_growth_form, crustose_morphology_growth_form,
         globular_morphology_growth_form, leathery_external_morphology,
         #articulated_external_morphology,
         branched_external_morphology, saccate_external_morphology,
         corticated_internal_anatomy,
         #physical_defenses_calcification,
         # canopy/subcanopy/turf
         space_use_canopy_gom, space_use_subcanopy_gom, space_use_turf_gom,
         # life history traits
         potential_asexual_reproduction, perennial_gom,
         #size_categories,
         # where/on what does it grow
         epilithic_env_position, 
         #epiphytic_env_position, epizoic_env_position,
         epibiotic_env_position_gom
         ) %>%
# Make everything a factor
  mutate(across(.fns = ~ as.factor(.)))

animal_raw <- animal_raw %>%
  select(species, phylum, common_name_division, # for filtering later
         # subtidal/intertidal for filtering later
         subtidal, intertidal,
         # basic traits
         solitary_colonial, adult_body_size_bin, 
         calcareous, motility_adult,
         # trophic traits
         herbivore, predator, omnivore, deposit_feeder,
         filter_suspension, scavenger, trophic_level,
         # habitat traits
         benthic, pelagic, epibiotic
         ) %>%
  # holdover from past script, not there's still whitespace issues tbh
  mutate(trophic_level = str_trim(trophic_level)) %>% 
  # make everything a factor
  mutate(across(.fns = ~ as.factor(.)))
```




## Impute with Mice



```{r mice}
## algae
# run mice on select columns
algae_imp <- mice(algae_raw[,9:ncol(algae_raw)], m=5, method = 'cart', print = F)
algae_imp2 <- complete(algae_imp) # necessary!
# reattach taxonomy/filtering columns
algae_imp2 <- cbind(algae_raw[,1:8],algae_imp2)

#to see logged events
algae_imp$loggedEvents

## animals
# run mice on select columns
animal_imp <- mice(animal_raw[,6:ncol(animal_raw)], m=5, method = 'cart', print = F)
animal_imp2 <- complete(animal_imp) #necessary!
# reattach taxonomy/filtering columns
animal_imp2 <- cbind(animal_raw[,1:5],animal_imp2)


#to see logged events
animal_imp$loggedEvents

```



```{r save}
# write csv for imputed data
write.csv(algae_imp2, file = here('data', 'clean','20220525_algae_imputed.csv'))
write.csv(animal_imp2, file = here('data', 'clean','20220321_animal_imputed.csv'))

# if you want .RData versions
#save(algae_imp2, file = here('data','algae_imputed.RData'))
#save(animal_imp2, file = here('data','animals_imputed.RData'))

```