Glance at an lm_sql_result

Description

Extract a single-row tibble of model-level summary statistics from a fitted SQL linear model.

Usage

## S3 method for class 'lm_sql_result'
glance(x, ...)

Arguments

Details

Returns R-squared, adjusted R-squared, residual standard error, F-statistic and its p-value, model degrees of freedom, log-likelihood, AIC, BIC, number of observations, and residual degrees of freedom.

Value

A single-row tibble with columns 'r.squared', 'adj.r.squared', 'sigma', 'statistic', 'p.value', 'df', 'logLik', 'AIC', 'BIC', 'nobs', and 'df.residual'.

SQL-Backed Linear Regression

Description

Fits a linear regression model using SQL aggregation on a remote database table. The data never leaves the database — only sufficient statistics (sums and cross-products) are returned to R.

Usage

lm_sql(formula, data, tol = 1e-07)

Arguments

Details

The function computes the X^TX and X^Ty matrices entirely inside the database engine via a single SQL aggregation query, then solves the normal equations in R using Cholesky decomposition (falling back to Moore-Penrose pseudoinverse for rank-deficient designs).

Supported formula features:

• Numeric and categorical (character/factor) predictors with automatic dummy encoding via 'CASE WHEN'.

• Interaction terms ('*' and ':') including numeric × categorical and categorical × categorical cross-products.

• Dot expansion ('y ~ .') to all non-response columns.

• Transforms: 'I()', 'log()', and 'sqrt()' translated to SQL equivalents ('POWER', 'LN', 'SQRT').

• Date and datetime predictors automatically cast to numeric in SQL.

• No-intercept models ('y ~ 0 + x').

For grouped data (via [dplyr::group_by()]), a single 'GROUP BY' query is executed and one model per group is returned in a tibble with a 'model' list-column.

NA handling uses listwise deletion: rows with 'NULL' in any model variable are excluded via a 'WHERE ... IS NOT NULL' clause.

Value

An S7 object of class lm_sql_result, or a tibble with a model list-column if the data is grouped.

Result object for SQL-backed Linear Model

Description

An S7 class that stores the complete results of a SQL-backed linear regression fitted by [lm_sql()].

Usage

lm_sql_result(
  coefficients = integer(0),
  std_error = integer(0),
  sigma = numeric(0),
  r_squared = numeric(0),
  adj_r_squared = numeric(0),
  f_statistic = integer(0),
  f_p_value = integer(0),
  logLik = numeric(0),
  AIC = numeric(0),
  BIC = numeric(0),
  nobs = numeric(0),
  df_residual = numeric(0),
  df_model = numeric(0),
  statistic = integer(0),
  p_value = integer(0),
  call = NULL,
  term_expressions = NULL
)

Details

This class is not called directly by users. It is created internally by [lm_sql()] and returned as the model object. It stores fitted coefficients, standard errors, t-statistics, p-values, and model-level summaries (R-squared, F-statistic, AIC, BIC, etc.). The 'term_expressions' property holds named R expressions for each predictor, which are used by the [orbital.lm_sql_result()] method to generate in-database prediction expressions.

Convert an lm_sql_result to an orbital object

Description

Creates an orbital object from a fitted SQL linear model, enabling in-database predictions without pulling data into R.

Usage

orbital.lm_sql_result(x, ..., prefix = ".pred")

Arguments

Details

Builds a single prediction expression by combining the fitted coefficients with the R expressions stored in 'term_expressions'. For categorical predictors, the expression includes 'ifelse()' calls that dbplyr translates to SQL 'CASE WHEN'. The resulting 'orbital_class' object can be used with [orbital::predict()] to get predictions or [orbital::augment()] to append a '.pred' column to a database table.

Value

An 'orbital_class' object.

Print an lm_sql_result

Description

Display a concise summary of a fitted SQL linear model.

Usage

## S3 method for class 'lm_sql_result'
print(x, ...)

Arguments

Details

Prints the original function call and the named coefficient vector.

Value

Invisibly returns 'x'.

Tidy an lm_sql_result

Description

Extract a tidy tibble of per-term coefficient statistics from a fitted SQL linear model.

Usage

## S3 method for class 'lm_sql_result'
tidy(x, conf.int = FALSE, conf.level = 0.95, ...)

Arguments

Details

Returns one row per model term with the estimate, standard error, t-statistic, and p-value. When 'conf.int = TRUE', confidence intervals are computed using the t-distribution with 'df_residual' degrees of freedom.

Value

A tibble with columns 'term', 'estimate', 'std.error', 'statistic', and 'p.value'. If 'conf.int = TRUE', also 'conf.low' and 'conf.high'.