MarketMatching (version 1.2.1)

roll_up_optimal_pairs: Roll up the suggested test/control optimal pairs for pseudo power analysis (testing fake lift)

Description

roll_up_optimal_pairs Takes the suggested optimal pairs from best_matches() and aggregates the data for pseudo power analysis (test_fake_lift()). You run this function and then pass the result (a matched markets object) to test_fake_lift.

Usage

roll_up_optimal_pairs(matched_markets=NULL,
                      percent_cutoff=1,
                      synthetic=FALSE)

Value

Returns an object of type market_matching. The object has the following elements:

BestMatches

A data.frame that contains the best matches for each market in the input dataset

Data

The raw data used to do the matching

MarketID

The name of the market identifier

MatchingMetric

The name of the matching variable

DateVariable

The name of the date variable

SuggestedTestControlSplits

Always NULL

Arguments

matched_markets

A matched market object from best_matches.

percent_cutoff

The percent of data (by volume) to be included in the future study. Default is 1. 0.5 would be 50 percent.

synthetic

If set to TRUE, the control markets are not aggregated so BSTS can determine weights for each market and create a synthetic control. If set to FALSE then the control markets are aggregated and each market will essentially get the same weight. If you have many control markets (say, more than 25) it is recommended to choose FALSE. Default is FALSE.

Examples

if (FALSE) {
##-----------------------------------------------------------------------
## Generate the suggested test/control pairs
##-----------------------------------------------------------------------
library(MarketMatching)
data(weather, package="MarketMatching")
mm <- best_matches(data=weather, 
                   id="Area",
                   date_variable="Date",
                   matching_variable="Mean_TemperatureF",
                   parallel=FALSE,
                   suggest_market_splits=TRUE,
                   start_match_period="2014-01-01",
                   end_match_period="2014-10-01")
                   
##-----------------------------------------------------------------------
## Roll up the pairs to generate test and control markets
## Synthetic=FALSE means that the control markets will be aggregated 
## -- i.e., equal weighhs in CausalImpact
##-----------------------------------------------------------------------
                   
rollup <- roll_up_optimal_pairs(matched_markets=mm, 
                                percent_cutoff=1, 
                                synthetic=FALSE)
                                
##-----------------------------------------------------------------------
## Pseudo power analysis (fake lift analysis)
##-----------------------------------------------------------------------

results <- test_fake_lift(matched_markets=rollup,
                     test_market="TEST",
                     lift_pattern_type="constant",
                     end_fake_post_period="2015-12-15",
                     prior_level_sd=0.002, 
                     max_fake_lift=0.1)
}