Skip to content





Create a version of the dictionary of results objects that uses less memory. This is the format of results dictionaries generated under the default options in EstimationModel (i.e. the following argument is executed: EstimationModel(..., full_results=False)). The SlimResults object is a smaller subset of the GLMResultsWrapper object in the statsmodels package(for more info, see statsmodels' GLMResults). Large attributes, such as copies of the estimating data, are removed from the results to cut back on memory size. The results most commonly referenced are retained, though.

The SlimResults object retains only the attributes listed below. For additional information see the documentation for the GLMResultsWrapper in the statsmodels package.


glm_results: statsmodels.genmod.generalized_linear_model.GLMResultsWrapper
  An instance of the results object


params: Pandas Series
  Estimated parameter values

aic: float
  Akaike Information Criterion

bic: float
  Bayes Information Criterion

llf: float
  Value of log-likelihood function

nobs: float
  number of observations

bse: Pandas Series
  Beta standard errors for parameter estimates

pvalues: Pandas Series
  Two-tailed pvalues for parameter estimates

family_name: str
  Name of distribution family used

family_link: str
  Estimation link function

method: str
  Estimation method

fit_history: int
  Number of iterations completed

scale: float
  The estimate of the scale / dispersion for the model fit

deviance: float
  Deviance measure

pearson_chi2: Pandas Series
  Chi-squared statistic

cov_type: str
  Covariance type

yname: str
  Column name of endogenous variable

xname: List[str]
  Column names of exogenous variables

model: str
  Model used for fit

df_resid: float

df_model: float

tvalues: Pandas Series
  T statistics

fittedvalues: Pandas Series
  Linear predicted values


The SlimResults object replicates two methods from the original GLMResultsWrapper object from statsmodels.

conf_int: array

create confidence intervals for parameter estimates. Arguments:

alpha: (optional) float
   The significance level for the confidence interval.
   I.e., The default alpha = .05 returns a 95% confidence interval.

cols: (optional) array-like
   cols specifies which confidence intervals to return

summary: object
  print a table summarizing estimation results (replicates statsmodels summary method   for GLM).