Skip to contents

Summarize an ordinary MARS fit

Usage

# S3 method for class 'mars'
summary(
  object,
  fit_measures = TRUE,
  digits = max(3, getOption("digits") - 3),
  vcov_method = NULL,
  inference = NULL,
  test = NULL,
  robust_type = NULL,
  hksj_adjust = NULL,
  cluster = NULL,
  level = 0.95,
  ...
)

Arguments

object

A fitted ordinary mars object.

fit_measures

Retained for compatibility.

digits

Display precision.

vcov_method

Optional "gls" or "observed" covariance override for this summary only. Set the choice in mars() to propagate it downstream.

inference

NULL preserves the usual summary; model or robust explicitly selects the coefficient inference framework for this summary only.

test

NULL uses normal model inference or the stored robust test. Options are z, t, Satterthwaite, and hksj (knha alias). Robust inference additionally accepts naive-t and naive-tp; robust t means naive-t (clusters minus one df).

robust_type

Optional robust covariance, such as CR2 or CR3. Defaults to the stored selection, or CR2 with an explicit cluster.

hksj_adjust

standard (default) or modified, which floors the Hartung-Knapp variance multiplier at one. Only valid with test='hksj'.

cluster

Optional column name in the fitted analysis data identifying independent robust clusters. Defaults to the stored robust clustering column.

level

Pointwise confidence level for location and scale coefficients, default 0.95.

...

Reserved; unused arguments are rejected.

Value

A native MARS summary including covariance method and optimizer status.

Details

Summary inference does not refit or modify the fitted object. The default retains the existing model and, where fitted, robust coefficient tables. An explicit selection returns one selected coefficient table, including df, and inference_details with its covariance, method and adjustment. Model t and HKSJ require independent univariate models without a scale formula. HKSJ additionally requires estimated heterogeneity and GLS covariance; its multiplier is the fitted weighted residual sum of squares divided by n-p. Model-based Satterthwaite uses conditional GLS covariance, covariance derivatives and inverse expected ML/REML covariance-parameter information: \(\mathrm{df}=2v^2/(g'I^{-1}g)\), where \(v\) is a coefficient variance and \(g\) is its gradient. It supports interior, identifiable ordinary univariate, multivariate and multilevel covariance models; location-scale and fixed-heterogeneity fits are excluded. This expected-information approximation is distinct from observed-Hessian Satterthwaite and from robust CR2/Satterthwaite. Robust overrides require independent clusters and never silently fall back from a requested covariance type. Unsupported combinations raise errors. Only mean-coefficient inference changes: existing heterogeneity, scale, prediction, joint model-fit tests and downstream analyses are unchanged. Bootstrap, jackknife and permutation computations remain explicit through mars_alt_estimation() and their own summary methods. Use the same options in confint.mars() for matching coefficient intervals.

References

Tipton, E. (2015). Small sample adjustments for robust variance estimation with meta-regression. Psychological Methods, 20(3), 375–393. doi:10.1037/met0000011 .

Pustejovsky, J. E., and Tipton, E. (2018). Small-sample methods for cluster-robust variance estimation and hypothesis testing in fixed effects models. Journal of Business & Economic Statistics, 36(4), 672–683. doi:10.1080/07350015.2016.1247004 .

Sidik, K., and Jonkman, J. N. (2002). A simple confidence interval for meta-analysis. Statistics in Medicine, 21(21), 3153–3159. doi:10.1002/sim.1262 .

Knapp, G., and Hartung, J. (2003). Improved tests for a random effects meta-regression with a single covariate. Statistics in Medicine, 22(17), 2693–2710. doi:10.1002/sim.1482 .

Röver, C., Knapp, G., and Friede, T. (2015). Hartung-Knapp-Sidik-Jonkman approach and its modification for random-effects meta-analysis with few studies. BMC Medical Research Methodology, 15, 99. doi:10.1186/s12874-015-0091-1 .

See also

mars_inference_options for the model compatibility table and supported argument combinations.