Compute Row-wise Means for Groups of Columns

Description

This function computes the row-wise mean for consecutive groups of columns in a matrix or data frame.

Usage

colsMean(x, ncols)

Arguments

Details

The function splits the columns of x into consecutive groups of ncols columns and calculates the mean of each row for each group. The number of columns in x must be divisible by ncols.

Value

A numeric matrix with the same number of rows as x and ncol(x) / ncols columns, where each column is the row-wise mean of a group of ncols columns.

Examples

mat <- matrix(1:12, nrow = 3)
colsMean(mat, 2)

Merge Soil Laboratory Data with Spectral Data

Description

This function merges soil laboratory data with cleaned spectral (VNIR) data, performs preprocessing, and prepares inputs for calibration and model building.

Usage

merge_of_lab_and_spectrum(soil_data, data_NaturaSpec_cleaned)

Arguments

Details

The function performs the following steps:

• Aggregates spectral data by wavelength and computes mean reflectance values.

• Merges the soil and spectral datasets by LAB_NUM.

• Separates soil variables and VNIR spectral matrix.

• Creates calibration sample indices using random sampling.

• Defines spectral bands to remove (detector artifact areas) and indices to be used in modeling.

Value

A list with the following elements:

soil

Data frame of soil laboratory data (first 8 columns of merged dataset).

vnir.matrix

Matrix of VNIR spectral reflectance values (without metadata columns).

j

List of calibration sample indices for cross-validation (4 sets).

rm1, rm2, rm3, rm4

Vectors of indices corresponding to spectral bands to be removed (detector artifact regions around 1000 nm and 1800 nm).

ind

Indices of spectral bands used for aggregation (columns 7–2146).

rmove

Indices of bands to be excluded from analysis.

vars

Vector of spectral band names retained after removal.

Examples

merged <- merge_of_lab_and_spectrum(soil_data, data_NaturaSpec_cleaned)
str(merged)

Machine Learning Function for Soil Spectral Data

Description

This function applies several machine learning models (PCR, PLSR, Random Forest, LASSO, Cubist) to soil spectral data and compares their performance. Optionally, it can return the best-performing model.

Usage

ml_f(
  x,
  y,
  smoother_selection,
  type_of_soil,
  model_selection = TRUE
)

Arguments

Details

The function merges spectral and laboratory data, preprocesses the data, and evaluates the following models:

• PCR (Principal Component Regression)

• PLSR (Partial Least Squares Regression)

• RF (Random Forest)

• LASSO regression

• Cubist regression

Each model's performance results are combined into a single results object. If model_selection = TRUE, the function returns the model with the highest performance metric (based on the 11th column of the results table).

Value

A data frame:

If model_selection = FALSE

Returns results for all models.

If model_selection = TRUE

Returns only the best-performing model result.

Examples

# Example usage:
results <- ml_f(
  x,
  y,
  smoother_selection = "savitzky",
  type_of_soil = "loam",
  model_selection = TRUE
)

Compute Model Evaluation Metrics

Description

This function computes various statistics for comparing observed values 'y' with predicted values 'yhat'. It includes correlation, regression coefficients, bias, RMSE, MSE, and predictive performance metrics like RPD and RPIQ.

Usage

msd.comp(y, yhat)

Arguments

Value

A named numeric vector with the following components:

r

Pearson correlation between 'y' and 'yhat'

int

Intercept of regression of 'y' on 'yhat'

slope

Slope of regression of 'y' on 'yhat'

r2

Coefficient of determination (R-squared)

bias

Mean bias: mean(yhat) - mean(y)

rmse

Root mean squared error

mse

Mean squared error

sb

Systematic bias component of MSE

nu

Non-unity slope component of MSE

lc

Lack-of-correlation component of MSE

rmse.c

Corrected RMSE after removing bias

mse.c

Corrected MSE after removing bias

rpd

Ratio of standard deviation to RMSE (RPD)

rpiq

Ratio of interquartile range to RMSE (RPIQ)

Examples

y_obs <- c(1.2, 3.4, 2.5, 4.1)
y_pred <- c(1.1, 3.5, 2.4, 4.0)
msd.comp(y_obs, y_pred)

Aggregate VNIR Spectra by Columns

Description

This function aggregates VNIR (Visible and Near-Infrared) spectral data by calculating the mean of every 10 columns while removing specific detector artifact regions (~1000nm and ~1800nm) and unwanted spectral bands.

Usage

raw(vnir.matrix)

Arguments

Details

The function removes columns corresponding to detector artifacts: - rm1: bands 1–46 - rm2: bands 637–666 - rm3: bands 1437–1466 - rm4: bands 2127–2151 Additionally, columns 1:4, 64:66, 144:146, and 213:214 (after averaging) are removed.

Value

A data frame containing the aggregated VNIR spectra with cleaned band names. The number of columns is reduced by averaging over every 10 bands and removing artifact-prone regions.

Examples

  raw_spectra <- raw(vnir_matrix)

Soil and Spectral Data Preprocessing for Model Training

Description

These functions fit predictive models for soil properties using VNIR spectral data. Each function applies a specific machine learning method:

• pcr_preprocess() – Principal Component Regression (PCR)

• plsr_preprocess() – Partial Least Squares Regression (PLSR)

• lasso_preprocess() – LASSO regression

• rf_preprocess() – Random Forest regression

• cubist_preprocess() – Cubist regression

Computes mean performance metrics across multiple calibration and validation sets. Typically used to summarize the results of soil property prediction models generated by preprocessing functions such as pcr_preprocess(), plsr_preprocess(), lasso_preprocess(), rf_preprocess(), or cubist_preprocess().

Usage

pcr_preprocess(soil, vnir.matrix, j, preprocess, type_of_soil)

plsr_preprocess(soil, vnir.matrix, j, preprocess, type_of_soil)

lasso_preprocess(soil, vnir.matrix, j, preprocess, type_of_soil)

rf_preprocess(soil, vnir.matrix, j, preprocess, type_of_soil)

cubist_preprocess(soil, vnir.matrix, j, preprocess, type_of_soil)

results(metric.list, soil_type)

Arguments

Details

All functions use the same workflow:

1. Combine the selected soil property with preprocessed spectra.

2. Split data into calibration and validation sets (using sample indices).

3. Fit the chosen model across multiple calibration/validation partitions.

4. Generate predictions and compute performance metrics (MSD-based).

Value

A list of MSD metric objects for calibration and validation sets, specific to the fitted model.

A named numeric vector of mean performance metrics across all splits:

LV

Latent variable / model index

cv-r2

Cross-validated R-squared for calibration set

cv-bias

Bias in cross-validation for calibration set

cv-rmse

Root mean squared error in cross-validation for calibration set

cal-mse

Mean squared error for calibration set

cal-rpiq

Ratio of performance to interquartile distance for calibration set

val-r2

R-squared for validation set

val-bias

Bias for validation set

val-rmse

Root mean squared error for validation set

val-mse

Mean squared error for validation set

val-rpiq

Ratio of performance to interquartile distance for validation set

See Also

merge_of_lab_and_spectrum, ml_f

Examples

# Example with PCR
results_pcr <- pcr_preprocess(soil, vnir.matrix, j, preprocess = scale, type_of_soil = 2)

# Example with Random Forest
results_rf <- rf_preprocess(soil, vnir.matrix, j, preprocess = scale, type_of_soil = 2)



msd_list <- pcr_preprocess(soil, vnir.matrix, j, preprocess = scale, type_of_soil = 2)
results_summary <- results(msd_list)