expanse_monthly

Version 0.1.0

A short description of your project

Project organization

``` . ├── .gitignore ├── CITATION.md ├── LICENSE.md ├── README.md ├── requirements.txt ├── bin <- Compiled and external code, ignored by git (PG) │ └── external <- Any external source code, ignored by git (RO) ├── config <- Configuration files (HW) ├── data <- All project data, ignored by git │ ├── processed <- The final, canonical data sets for modeling. (PG) │ ├── raw <- The original, immutable data dump. (RO) │ └── temp <- Intermediate data that has been transformed. (PG) ├── docs <- Documentation notebook for users (HW) │ ├── manuscript <- Manuscript source, e.g., LaTeX, Markdown, etc. (HW) │   └── reports <- Other project reports and notebooks (e.g. Jupyter, .Rmd) (HW) ├── results │ ├── figures <- Figures for the manuscript or reports (PG) │   └── output <- Other output for the manuscript or reports (PG) └── src <- Source code for this project (HW)

```

Source code

• 01model5foldmonthlysep.R trains and evaluates LUR models for every month using 5-fold cross-validation. It uses two algorithms to train the LUR models: supervised linear regression (SLR) and random forests (RF).
• 01model5foldmonthlysep_gwr.R trains LUR models for every month using 5-fold cross-validation. It uses geographically weighted regression (GWR) to train the LUR models.

• 01outputpredictiontestcv.R combines all hold-out validation data from the 5-fold CV.

• 02modelallmonthlysep.R trains LUR models for every month using all available observations. It uses two algorithms to train the LUR models: supervised linear regression (SLR) and random forests (RF).

• 02modelallmonthlysep_gwr.R trains LUR models for every month using all available observations. It uses geographically weighted regression (GWR) to train the LUR models. -- 02visrandomPoints_IDW.R: processes and combines csv files containing estimates from several models at random points into one rds file (from predictionsAllmonthlyRandomidwAndannualComparisonPOLLblockXXXXX.csv to randomPointsAll_cleaned.rds). It also creates heatmap plot of indicating the correlation values between the model estimates at these random points.

Function code

• 00funreaedmonthlydata_gee.R contains the 'read_data' function that reads monthly predictors for each pollutant and year(s).
• funoutputrf.R outputs RF predictions and variable importance as csv files.
• funrfrf.R trains RF models.

Data

• Files /data/raw/gee/predictionsAllmonthlyRandomidwAndannualComparisonPOLLblockXXXXX.csv contain the estimates from different models (annual GTWR, monthly GWR, monthly adjusted GTWR using two inverse distance weighting methods) at random points.
• The processed and combined file from the abovementioned files is data/processed/randomPointsAll_cleaned.rds

License

This project is licensed under the terms of the MIT License

Citation

Please cite this project as described here.