Matching Refinement with Devontez Data: Real Identities vs. V1 and V2 Matching

Purpose/Context:

Compare person-centric metrics as generated by real identities in the golden dataset to person-centric metrics generated by reliance on v1 and v2 matching algorithms.

A note on terminology/methods/decisions:

• A “person” is determined from the following fields:
  • Real.Identity (in the context of interpreting "Real Identity" from raw golden data)
  • Worker.ID (in the context of interpreting v1 matching efforts from raw golden data)
  • person_id (in the context of interpreting v2 matching efforts from newly processed data)

• Because each stat is person-centric, observations (ads) that have no Real.Identity/Worker.ID/person_id are not considered.

• "Ad counts across persons" considers all ads associated with a person, not just unique ads.

• Both mean and median are provided. Friendly reminder that the median is a better measurement of central tendency for data with skewed distributions (which this data has), as the median is more robust/not as affected by outliers.

• Age-related stats omit ads with no associated age.

Persons Per Matching

Ads Per Person

Real Identity

##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
##     3.0    16.0    94.5   114.6   198.2   308.0

V1 Matching

##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
##     1.0     4.0    13.0    57.7    43.0  1103.0

V2 Matching

##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
##    1.00    1.00    5.00   83.24   22.50 2035.00

Cities Per Person

Real Identity

##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
##   1.000   2.000   4.000   9.286  11.250  42.000

V1 Matching

##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
##   1.000   1.000   2.000   4.034   4.000  66.000

V2 Matching

##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
##   1.000   1.000   1.000   4.445   3.000  85.000

States Per Person

Real Identity

##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
##   1.000   1.250   2.500   5.286   6.500  19.000

V1 Matching

##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
##   1.000   1.000   1.000   2.298   2.000  21.000

V2 Matching

##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
##   1.000   1.000   1.000   2.218   2.000  31.000

Phones Per Person

Real Identity

##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
##   1.000   4.000   5.000   9.786   8.500  40.000

V1 Matching

##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
##   1.000   1.000   3.000   4.226   5.000  53.000

V2 Matching

##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
##   1.000   1.000   1.000   4.983   3.000 161.000

Days Lapsed Across Ads Per Person

Real Identity

##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
##    14.0   185.5   230.0   377.3   526.2  1195.0

V1 Matching

##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
##    0.00    6.75   54.00  115.81  172.25  757.00

V2 Matching

##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
##    0.00    0.00    6.00   84.84   76.00  772.00

Ages Per Person

Real Identity

##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
##   1.000   2.000   2.000   3.357   4.000   8.000

V1 Matching

##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
##   1.000   1.000   2.000   2.274   3.000  12.000

V2 Matching

##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
##   1.000   1.000   1.000   1.916   2.000  12.000

Age Ranges Per Person

Real Identity

##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
##   0.000   1.000   2.500   4.214   4.750  26.000

V1 Matching

##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
##   0.000   0.000   1.000   2.625   3.000  30.000

V2 Matching

##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
##    0.00    0.00    0.00    3.05    2.00   30.00