Struct LinearRegression

Source

pub struct LinearRegression<T = f64>where
    T: Float + Debug + Default + Serialize,{
    pub slope: T,
    pub intercept: T,
    pub r_squared: T,
    pub standard_error: T,
    pub n: usize,
}

Expand description

Linear regression model that fits a line to data points.

Fields§

§slope: T

Slope of the regression line (coefficient of x)

§intercept: T

Y-intercept of the regression line

§r_squared: T

Coefficient of determination (R²) - goodness of fit

§standard_error: T

Standard error of the estimate

§n: usize

Number of data points used for regression

Implementations§

Source §

impl<T> LinearRegression<T>
where T: Float + Debug + Default + NumCast + Serialize + for<'de> Deserialize<'de>,

Source

pub fn new() -> Self

Create a new linear regression model without fitting any data

Source

pub fn fit<U, V>(&mut self, x_values: &[U], y_values: &[V]) -> StatsResult<()>
where U: NumCast + Copy, V: NumCast + Copy,

Fit a linear model to the provided x and y data points

§Arguments

x_values - Independent variable values
y_values - Dependent variable values (observations)

§Returns

StatsResult<()> - Ok if successful, Err with StatsError if the inputs are invalid

§Errors

Returns StatsError::DimensionMismatch if X and Y arrays have different lengths. Returns StatsError::EmptyData if the input arrays are empty. Returns StatsError::ConversionError if value conversion fails. Returns StatsError::InvalidParameter if there’s no variance in X values.

Source

pub fn predict(&self, x: U) -> StatsResult<T>
where U: NumCast + Copy,

Predict y value for a given x using the fitted model

§Arguments

x - The x value to predict for

§Returns

StatsResult<T> - The predicted y value

§Errors

Returns StatsError::NotFitted if the model has not been fitted (n == 0). Returns StatsError::ConversionError if type conversion fails.

§Examples

use rs_stats::regression::linear_regression::LinearRegression;

let mut model = LinearRegression::<f64>::new();
model.fit(&[1.0, 2.0, 3.0], &[2.0, 4.0, 6.0]).unwrap();

let prediction = model.predict(4.0).unwrap();
assert!((prediction - 8.0).abs() < 1e-10);

Source

pub fn predict_many(&self, x_values: &[U]) -> StatsResult<Vec<T>>
where U: NumCast + Copy + Send + Sync, T: Send + Sync,

Calculate predictions for multiple x values

§Arguments

x_values - Slice of x values to predict for

§Returns

StatsResult<Vec<T>> - Vector of predicted y values

§Errors

Returns StatsError::NotFitted if the model has not been fitted. Returns StatsError::ConversionError if type conversion fails for any value.

§Examples

use rs_stats::regression::linear_regression::LinearRegression;

let mut model = LinearRegression::<f64>::new();
model.fit(&[1.0, 2.0, 3.0], &[2.0, 4.0, 6.0]).unwrap();

let predictions = model.predict_many(&[4.0, 5.0]).unwrap();
assert_eq!(predictions.len(), 2);

Source

pub fn confidence_interval( &self, x: U, confidence_level: f64, ) -> StatsResult<(T, T)>
where U: NumCast + Copy,

Calculate confidence intervals for the regression line

§Arguments

x - The x value to calculate confidence interval for
confidence_level - Confidence level (0.95 for 95% confidence)

§Returns

StatsResult<(T, T)> - Tuple of (lower_bound, upper_bound), or an error if invalid

§Errors

Returns StatsError::InvalidInput if there are fewer than 3 data points. Returns StatsError::InvalidParameter if confidence level is not supported (only 0.90, 0.95, 0.99). Returns StatsError::ConversionError if value conversion fails.

Source

pub fn correlation_coefficient(&self) -> StatsResult<T>

Get the correlation coefficient (r)

The correlation coefficient ranges from -1 to 1, indicating the strength and direction of the linear relationship between x and y.

§Returns

StatsResult<T> - The correlation coefficient

§Errors

Returns StatsError::NotFitted if the model has not been fitted (n == 0).

§Examples

use rs_stats::regression::linear_regression::LinearRegression;

let mut model = LinearRegression::<f64>::new();
model.fit(&[1.0, 2.0, 3.0], &[2.0, 4.0, 6.0]).unwrap();

let r = model.correlation_coefficient().unwrap();
assert!((r - 1.0).abs() < 1e-10); // Perfect positive correlation