| output | title | author |
|---|---|---|
html_document |
ReadMe |
Vanessa Haller-Bull |
collection and computations to prepare data for the species choice analysis
The data for the traits of concern is downloaded from www.coraltraits.org. For quantitative traits we download the average +/- standard deviation.
The main code to complete the data download is "DataPrep.R", it utilises the inputs on species of concern as well as their taxonomy provided by Josh.
The output is saves as "SpeciesData.csv" and contains the following traits:
Table 1: Number of missing values for the data downloaded
| Trait | Missing Data |
|---|---|
| skeletaldensityMin | 382 |
| skeletaldensityMax | 382 |
| growthrateMin | 357 |
| growthrateMax | 357 |
| corallitewidthMin | 50 |
| corallitewidthMax | 20 |
| colonydiameterMin | 254 |
| colonydiameterMax | 254 |
| polypfecundityMin | 403 |
| polypfecundityMax | 403 |
| eggsizeMin | 411 |
| eggsizeMax | 411 |
| photosynthesisMin | 408 |
| photosynthesisMax | 408 |
| BRI | 202 |
Next, we download the extra data that will be used to impute missing datapoints for each trait. This is done through the code called "[Trait name]Data.R" and teh results are saved as "[Trait name]BeforeImputed.RData".
The following traits are used to support each of the traits in the main analysis.
Table 2: Traits used for the imputation for each variable in the final dataset
| Variable in final dataset | Variables used for imputation |
|---|---|
| Skeletal density | Larval swimming speed, Colony shape factor, Substrate attachment, Wave exposure, Wave exposure preference |
| Growth rate | Calcification rate, Life history strategy, Growth form, Growth form from Veron |
| Corallite width | Polyps per area, Tissue thickness, Total biomass |
| Colony diameter | Colony area, Colony shape factor, Coloniality |
| Reproduction | Polyp fecundity, Egg size, Colony fecundity, Mode of larval development, Propagule size, Sexual system, Symbiodinium |
| Photosynthesis | Symbiodinium density, Symbiodinium clade, Symbiodinium subclade, Zooxanthellate |
The imputation trials different methods and repeats these multiple times to receive teh best results. The code is saved as "ImputeMissingData[Trait name].R" and the results are saved in multiple dataframes named "[Trait name]ImputedData[Numbering].RData".
To test the imputation results, I iteratively remove one known datapoint and refit the model to this dataset and then predict this removed datapoint using the fitted model. Then I calculate the mean +/- standard deviation for each of these predicted datapoints which can then be compared to the known datapoint prior to removal.
The code to conduct this analysis is named "CheckImputedData[Trait name].R".