Skip to main content

Table 1 Covariates balance before and after matching, Milan, Italy, 2003–2006

From: Assessing the short term impact of air pollution on mortality: a matching approach

  Mean/Proportion Standardized differenced
Background characteristic Treated (n = 812) Controls (n = 649) Matched Controls (n = 649) Pre-matching Post-matching % Biase
Estimated propensity score 0.756 0.306 0.756 1.810 0 100.0
Temperature (°C)a 11.4 18.3 11.3 0.914 0.013 98.5
Relative humidity (%) 66.8 58.6 67.1 0.456 0.014 97.0
Saturdays and Sunday 0.243 0.341 0.195 0.217 0.106 51.0
Day of year - - - 405.5c 15.9c 96.1f
Influenza epidemics 0.128 0.009 0.054 0.483 0.315 34.7
Heat episodesb 0.032 0.028 0.025 0.001 0.002 −77.8
Summer days 0.225 0.664 0.252 0.037 0.002 93.8
  1. aTemperature: average temperature in the current and in the previous 3 days
  2. bHeat episodes: days with temperatures exceeding 28 °C. Summer days: from May 1st to September 30th
  3. cWheeler and Watson’s statistics (W) for the comparison of the distribution of the circular variable “day of the year” between treated and control units (pre-matching W) or between treated and matched controls (post-matching W)
  4. dStandardized difference: pre-matching (\( {\delta}^{pre} \)) and post-matching (\( {\delta}^{post} \))
  5. e% bias: \( 100\times \left({\delta}^{pre}-{\delta}^{post}\right)/{\delta}^{pre} \)
  6. fPercent reduction in the Wheeler and Watson’s statistics after matching