Takes in all of the filled diff df CSV files and uses them to compute group level ATTs as well as the aggregate ATT and its standard errors and p-values.
Usage
undid_stage_three(
dir_path,
agg = "silo",
weights = TRUE,
covariates = FALSE,
interpolation = FALSE,
save_csv = FALSE,
filename = "UNDID_results.csv",
filepath = tempdir(),
nperm = 1001,
verbose = TRUE
)Arguments
- dir_path
A character specifying the filepath to the folder containing all of the filled diff df CSV files.
- agg
A character which specifies the aggregation methodology for computing the aggregate ATT in the case of staggered adoption. Options are:
"silo","g", or"gt". Defaults to"silo".- weights
A logical value (either
TRUEorFALSE) which determines whether or not the weights should be used in the case of common adoption. Defaults toTRUE.- covariates
A logical value (either
TRUEorFALSE) which specifies whether to use thediff_estimatecolumn or thediff_estimate_covariatescolumn from the filled diff df CSV files when computing ATTs.- interpolation
A logical value or a character which specifies which, if any, method of interpolation/extrapolation for missing values of
diff_estimateordiff_estimate_covariatesshould be used. There must be at least onediff_estimateordiff_estimate_covariatesvalue for the (silo,g) group for which a missing value is being estimated in order for interpolation to work. Options are:"linear_function","nearest_value", or"piecewise_linear". Defaults toFALSE.- save_csv
A logical value, either
TRUEorFALSE(default), which determines if a CSV copy of the UNDID results will be saved or not.- filename
A string filename for the created CSV file. Defaults to
"UNDID_results.csv"- filepath
Filepath to save the CSV file. Defaults to
tempdir().- nperm
Number of random permutations of gvar & silo pairs to consider when calculating the randomization inference p-value. Defaults to
1001.- verbose
A logical value (either
TRUEorFALSE) which toggles messages showing the progress of the randomization inference. Defaults toTRUE.
Value
A data frame containing the aggregate ATT and its
standard errors and p-values from two-sided tests of agg_ATT == 0.
Also returns group (silo, g, or gt) level ATTs for staggered adoption.
Details
The agg parameter specifies the aggregation method used in the
case of staggered adoption. By default it is set to "silo" so that the ATTs
are aggregated across silos with each silo having equal weight, but can be
set to "gt" or "g" instead. Aggregating across "g" calculates ATTs for
groups based on when the treatment time was, with each "g" group having
equal weight. Aggregating across "gt" calculates ATTs for groups based on
when the treatment time was and the time for which the ATT is calculated.
The agg parameter is ignored in the case of a common treatment time and
only takes effect in the case of staggered adoption. For common adoption,
refer to the weights parameter.
Examples
# Execute `undid_stage_three()`
dir <- system.file("extdata/staggered", package = "undidR")
undid_stage_three(dir, agg = "g", nperm = 501, verbose = FALSE)
#> gvar gvar_ATT gvar_SE gvar_p_val gvar_jknife_SE
#> 1 1991-01-01 0.03824003 0.02557435 0.1394760603 0.02756434
#> 2 1993-01-01 0.03312283 0.02541522 0.1980131279 0.02691780
#> 3 1996-01-01 0.09702872 0.03208821 0.0048034212 0.03656550
#> 4 1997-01-01 0.11819478 0.02866265 0.0002715445 0.03064223
#> 5 1998-01-01 -0.05942719 0.05037900 0.2527181863 0.06848128
#> gvar_jknife_p_val agg_ATT agg_ATT_SE agg_ATT_p_val agg_ATT_jknife_SE
#> 1 0.1698802214 0.04543183 0.02768056 0.1760813 0.03094781
#> 2 0.2238366570 NA NA NA NA
#> 3 0.0121558950 NA NA NA NA
#> 4 0.0005644115 NA NA NA NA
#> 5 0.3963391606 NA NA NA NA
#> agg_ATT_jknife_p_val agg_ATT_RI_p_val
#> 1 0.2384033 0.214
#> 2 NA NA
#> 3 NA NA
#> 4 NA NA
#> 5 NA NA