--no_standardize

Switch

--no_standardize

Description

Disables column wise z-scoring of outcomes/features for regression/classification.

Argument and Default Value

False (i.e. standardizing is the default)

Details

Usually, every outcome is z:doc:fwflag_scored, and so are the group_norms for every feature, but this switch disables that. This can sometimes improve prediction performance though usually it's slightly worse than with standardizing.

--regression_to_lexicon, --classification_to_lexicon need this flag.

Other Switches

Required Switches: -d, -g, -t, -f, --outcome_table, --outcomes Won't do anything without any of these switches: --train_regression, --nfold_test_regression, etc. --train_classifiers, --nfold_test_classifiers, etc. --regression_to_lexicon --classification_to_lexicon Example Commands ================

Using the packaged dla_tutorial dataset (see Differential Language Analysis (DLA) Tutorial): trains a regression model to predict age for users from 1grams, without standardizing. Saves the model to a picklefile called age.pickle, and creates a lexicon called testAgeLex.

dlatkInterface.py -d dla_tutorial -t msgs -c user_id -f 'feat$1gram$msgs$user_id$16to16$0_01' \
--outcome_table blog_outcomes --outcomes age --train_regression --save_model --picklefile age.pickle \
--no_standardize --regression_to_lexicon testAgeLex