search

insongkim / PanelMatch

111 stars 34 forks source link

Missing values in propensity score #100

Closed ellenjunghyunkim closed 2 years ago

ellenjunghyunkim commented 2 years ago

Thank you again for your nice package!

How does ps.weight distribute weights to controlled units with missing values in some covariates? For example, the covariate tradewb has missing values in some units in your dataset (dem). It seems that the default setting in glm for the propensity score estimation is deleting the rows.(na.omit) However, we can still retrieve weights for units with missing values. Could you explain the weight allocation for the units with missingness in the covariates? Thank you so much!

Kind regards, Jung Hyun

ellenjunghyunkim commented 2 years ago

I found your answer from https://rdrr.io/cran/PanelMatch/src/R/pm_helpers_r.R # In practice, this function just looks over the data in the specified columns in the "data" data frame for missing data. Then it creates columns with indicator variables about the missingness of those variables I will close this issue.