diff --git a/chapters/10-estimands.qmd b/chapters/10-estimands.qmd index cee62ed..805cfc3 100644 --- a/chapters/10-estimands.qmd +++ b/chapters/10-estimands.qmd @@ -1328,45 +1328,49 @@ Since other estimands besides the ATE weaken causal assumptions, they might be m Below is a table summarizing the estimands and methods for estimating them (including R functions), along with sample questions extended from Table 2 in @greifer2021choosing. - +----------+----------------------------------+------------------------------------------------------------------------------------------------------------------------------------------------------+--------------------------------------+------------------+ - | Estimand | Target population | Example Research Question | Matching Method | Weighting Method | - +==========+==================================+======================================================================================================================================================+======================================+==================+ - | ATE | Full population | Should we have Extra Magic Hours *all* mornings to change the wait time for Seven Dwarfs Mine Train between 5–6 P.M.? | Full matching | ATE Weights | - | | | | | | - | | | Should a specific policy be applied to all eligible observations? | Fine stratification | `wt_ate()` | - | | | | | | - | | | | `matchit(…, estimand = "ATE")` | | - +----------+----------------------------------+------------------------------------------------------------------------------------------------------------------------------------------------------+--------------------------------------+------------------+ - | ATT | Exposed (treated) observations | Should we stop Extra Magic Hours to change the wait time for Seven Dwarfs Mine Train between 5–6 P.M.? | Pair matching without a caliper | ATT weights | - | | | | | | - | | | Should we stop our marketing campaign to those currently receiving it? | Full matching | `wt_att()` | - | | | | | | - | | | Should medical providers stop recommending treatment for those currently receiving it? | Fine stratification | | - | | | | | | - | | | | `matchit(…, estimand = "ATT")` | | - +----------+----------------------------------+------------------------------------------------------------------------------------------------------------------------------------------------------+--------------------------------------+------------------+ - | ATU | Unexposed (control) observations | Should we add Extra Magic Hours for all days to change the wait time for Seven Dwarfs Mine Train between 5–6 P.M.? | Pair matching without a caliper | ATU weights | - | | | | | | - | | | Should we extend our marketing campaign to those not receiving it? | Full matching | `wt_atu()` | - | | | | | | - | | | Should medical providers extend treatment to those not currently receiving it? | Fine stratification | | - | | | | | | - | | | | `matchit(…, estimand = "ATC")` | | - +----------+----------------------------------+------------------------------------------------------------------------------------------------------------------------------------------------------+--------------------------------------+------------------+ - | ATM | Evenly matchable | Are there some days we should change whether we are offering Extra Magic Hours to change the wait time for Seven Dwarfs Mine Train between 5–6 P.M.? | Caliper matching | ATM weights | - | | | | | | - | | | Is there an effect of the exposure for some observations? | `matchit(…, caliper = 0.1)` | `wt_atm()` | - | | | | | | - | | | Should those at clinical equipoise receive treatment? | Coarsened exact matching | | - | | | | | | - | | | | Cardinality matching | | - | | | | | | - | | | | `matchit(…, method = "cardinality")` | | - +----------+----------------------------------+------------------------------------------------------------------------------------------------------------------------------------------------------+--------------------------------------+------------------+ - | ATO | Overlap population | Same as ATM | | ATO weights | - | | | | | | - | | | | | `wt_ato()` | - +----------+----------------------------------+------------------------------------------------------------------------------------------------------------------------------------------------------+--------------------------------------+------------------+ + + ++----------+----------------------------------+------------------------------------------------------------------------------------------------------------------------------------------------------+--------------------------------------+------------------+ +| Estimand | Target population | Example Research Question | Matching Method | Weighting Method | ++==========+==================================+======================================================================================================================================================+======================================+==================+ +| ATE | Full population | Should we have Extra Magic Hours *all* mornings to change the wait time for Seven Dwarfs Mine Train between 5–6 P.M.? | Full matching | ATE Weights | +| | | | | | +| | | Should a specific policy be applied to all eligible observations? | Fine stratification | `wt_ate()` | +| | | | | | +| | | | `matchit(…, estimand = "ATE")` | | ++----------+----------------------------------+------------------------------------------------------------------------------------------------------------------------------------------------------+--------------------------------------+------------------+ +| ATT | Exposed (treated) observations | Should we stop Extra Magic Hours to change the wait time for Seven Dwarfs Mine Train between 5–6 P.M.? | Pair matching without a caliper | ATT weights | +| | | | | | +| | | Should we stop our marketing campaign to those currently receiving it? | Full matching | `wt_att()` | +| | | | | | +| | | Should medical providers stop recommending treatment for those currently receiving it? | Fine stratification | | +| | | | | | +| | | | `matchit(…, estimand = "ATT")` | | ++----------+----------------------------------+------------------------------------------------------------------------------------------------------------------------------------------------------+--------------------------------------+------------------+ +| ATU | Unexposed (control) observations | Should we add Extra Magic Hours for all days to change the wait time for Seven Dwarfs Mine Train between 5–6 P.M.? | Pair matching without a caliper | ATU weights | +| | | | | | +| | | Should we extend our marketing campaign to those not receiving it? | Full matching | `wt_atu()` | +| | | | | | +| | | Should medical providers extend treatment to those not currently receiving it? | Fine stratification | | +| | | | | | +| | | | `matchit(…, estimand = "ATC")` | | ++----------+----------------------------------+------------------------------------------------------------------------------------------------------------------------------------------------------+--------------------------------------+------------------+ +| ATM | Evenly matchable | Are there some days we should change whether we are offering Extra Magic Hours to change the wait time for Seven Dwarfs Mine Train between 5–6 P.M.? | Caliper matching | ATM weights | +| | | | | | +| | | Is there an effect of the exposure for some observations? | `matchit(…, caliper = 0.1)` | `wt_atm()` | +| | | | | | +| | | Should those at clinical equipoise receive treatment? | Coarsened exact matching | | +| | | | | | +| | | | Cardinality matching | | +| | | | | | +| | | | `matchit(…, method = "cardinality")` | | ++----------+----------------------------------+------------------------------------------------------------------------------------------------------------------------------------------------------+--------------------------------------+------------------+ +| ATO | Overlap population | Same as ATM | | ATO weights | +| | | | | | +| | | | | `wt_ato()` | ++----------+----------------------------------+------------------------------------------------------------------------------------------------------------------------------------------------------+--------------------------------------+------------------+ + + ### Effective Sample Size