Prediction Maps & Validation using Logistic Regression & ROC

Create dichotomous training data by reclassifying slope into safe and danger zones, selecting zones nine and ten as safe, and generating equal random points for logistic regression and roc validation.

Merge binary dichotomous training data into a single roster, ensuring both parties use the same system, then export and prepare the unified dataset for modeling and validation.

Explore how to set up the R studio project environment, create a working directory with output and libraries, clean the console, and organize code and data for reproducible analysis.

Learn how to install multiple required packages for the prediction maps task, including adding libraries, reading dependencies, and using code to install several packages.

Export raster factors from ArcGIS into R Studio, ensuring matching projection, cell size, and grid alignment; prepare a working directory and training data to enable reliable prediction modeling.

Load raster elevation data in R, create the innovation variable, and plot its histogram while reviewing the projection; multiply the elevation by five to compare changes.

Explore raster visualization and exporting in R by using the image function to control colors, set value ranges, and save elevation data as raster outputs.

Classify raster factors in R by building a three-column matrix, applying a vector function, and reclassifying innovation using defined metrics to produce categorized maps.

The lecture explains how to create color bars and breaks in R, adjusting breaks to reflect numerical locations across the full range, and selecting colors, greyscale, and symbology.

Learn to crop a specific raster area in R by defining an extent with a box, applying the crop function to elevation data, and handling extents precisely for targeted analysis.

Convert dependent and independent factors for a logistic regression model in R, using slope, elevation, curvature, and cover, and compare results across data extents.

Incorporate land cover and soil types as categorical factors in logistic regression for hazard mapping. Convert these classes to continuous representations to uncover relationships with landslides using common training data.

Convert categorical data into a numeric independent factor for a logistic regression model by counting, restructuring, and applying a consistent preparation procedure.

Create a training resample to align the training data with the data extent. Check the training and save the resampled data to your working directory.

Learn to read and stack raster data for logistic regression, align training data, and combine independent and dependent factors into a single file for modeling.

Identify and remove no-data entries from the roster data, ensuring numeric values; export the cleaned value table to a text file for 150 observations and prepare for a statistical test.

Run the logistic regression function using the training data, examine coefficients and intercept, assess curvature and significance, and review estimated standard errors and iteration counts for model discrimination.

Explore running ANOVA and McFadden R-squared tests, inspect regression results with coefficients, and discuss correlation and confidence intervals using a practical package.

Explore the confusion matrix in R to compare predicted versus actual values, inspect residuals and bias, and assess model accuracy against training data.

Learn to calculate and plot the AUC of the ROC curve to validate a predictive model, compare predicted versus actual data, and interpret positive and false positive rates.

Visualize how elevation and other predictors shape the logistic regression probability curve, compare training data, and interpret negative and positive relationships using graphs.

apply logistic regression coefficients in R studio to generate a prediction index map, converting linear predictors into probabilities and visualizing susceptibility via the final prediction map.

Translate logistic regression coefficients into a prediction index map in arcgis, compare model accuracy and validation against training and testing data, and discuss methods for improving precision.