PSYC 559 2nd In Class Quiz

Feature engineering

process of creating, transforming, or selecting variables (features) from raw data to improve a model’s performance

Missing Completely at Random (MCAR)

probability of a value being missing is unrelated to the value itself or any other observed or unobserved variable

purely random process (cat walking over keyboard)

note) dropping these cases will not bias the results

Missing at random (MAR)

probability of a value being missing is systematically related to other observed variables in the dataset, but not to the observed value itself. methods like multiple imputation can use income to predict and impute missing happiness values without bias. ex) missingness in happiness depends on other observed variables but not on happiness itself.

Missing not at random (MNAR)

missingness in happiness depends on the unobserved happiness score itself.

standard imputation methods will be biased. handling mnar requires explicitly modeling the missingness mechanism (e.g. selection models, pattern-mixture models)

Handling Missing Data in R

Recoding missing values as NA.

Listwise Deletion

Imputation

Missing Data Visualization

once data has been recoded, visualizations used to explore the amount of missing data in dataset.

Listwise Deletion

also called complete-case analysis

remove any row that has one or more missing values across any variable used in the analysis

should only be used if data is missing completely at random (MCAR). can result in less efficiency (reduced power and more uncertainty)

Imputation

process of filling in missing values in a dataset with estimated or predicted values so that you can perform analyses without dropping incomplete cases