Professional Documents
Culture Documents
Supplement 1 MeasuringOutcomesandImp
Supplement 1 MeasuringOutcomesandImp
Policy Intervention
8 6 4
2 0 1 2 3 4 5 6 7 8 9
Sequence number
Activity theory of causation Recognition of systemic complexity Financial, political, and ethical feasibility Rarity of true experiments Availability of resources (www.economagic.com, www.fedstats.gov, www.census.gov, www.eurostat, Statisticki Godisnjak/ Bilten)
Control Series
I II O1 O1 O2 O2 O3 O3 X ~X O4 O4 O5 O5 O6 O6 O7 O7
Policy Intervention
8 6 4
2 0 1 2 3 4 5 6 7 8 9
Sequence number
HousingArea per person (square meters) Average Life Expectancy Quality Adjusted Life Years Persons Below Poverty Line Income Distribution (Gini Index) Air Pollution Index (parts per million) Lead Concentration Index (blood concentration)
797 (0.48)
1,654 (1.0)
P / 1-P = 0.92
111 (0.58)
1-Q
82 (0.42)
193 (1.0)
Q / 1-Q = 1.38
texp = mean fatality rate after the policy - mean fatality rate before the policy / pooled standard deviation = -0.23 / 0.35 = -0.66 tcon = mean fatality rate after the policy mean fatality rate before the policy / pooled standard deviation = -0.07 / 0.10 = -0.70
0.80-0.99 strong 0.60-0.79 moderate to strong 0.40-0.59 moderate 0.20-0.39 weak to moderate 0.00-0.19 negligible to weak
NOTE: The practical significance of an effect size depends on the social costs of being wrong.
Pooled t-Test. The outcome mean after the intervention subtracted from the outcome mean before the intervention, divided by the pooled standard deviation. NOTE: The observations before and after the intervention are not independent and therefore the pooed t-test must be used. x2 x1 / sqrt [Sp (1/n1) + (1/n2)] Standard (z) Scores. An individual score subtracted from the mean of the distribution divided by the standard deviation. z = x - m / s . An individual score of 116 from a distribution with a mean of 100 and a standard deviation of 16 is the same as an individual score of 348 from a distribution with a mean of 300 and a standard deviation of 48. Individual scores measured with two different scales, or from two different distributions, can be directly compared.
320
FATALITIES
310
300
290
280 1954
1955
1956
1957
YEAR
52000
52000
FATALITIES
50000
50000
48000
48000
46000
46000
YEAR
FATALITIES
300
280
260
240 220 . 1951 1952 1953 1954 1955 1956 1957 1958 1959 .
YEAR
FATALITIES
500 00
400 00
300 00 196 6 197 0 197 4 197 8 198 2 198 6 199 0 199 4 199 8 200 0 196 8 197 2 197 6 198 0 198 4 198 8 199 2 199 6
YEAR
320
300
FATALITIES
280
260
240
220
YEAR
0.0
-.1 Fatalities -.2 1968 1972 1976 1980 1984 1988 1992 1996 2000 1970 1974 1978 1982 1986 1990 1994 1998 Economic Index
YEAR
Transforms: natural log, difference (1)
24
22
20 US 18 EURCON 16 70 71 72 73 74 75 76 EUREXP
YEAR
0.0
-.2
19
96 19 0 9 19 4 8 19 8 7 19 2 7 19 6 6 19 0 6 19 4 5 19 8 4 19 2 4 19 6 3 19 0 3 19 4 2 19 8 1 19 2 1 19 6 0
Year
Group Problem
Examine the extended time-series graphs showing the observed fatality rate, the predicted fatality rate, and European Commission target for 2010.
1. Explain how interrupted time-series analysis might result in a different predicted fatality rate. Is the observed fatality rate a valid predictor of fatalities in future?
2. Explain how control-series analysis might change the Commissions 2010 target fatality rate. Is the target realistic?
40 30 20 10 0 1980
1985
1995 model
2000
2005
2010
ETSC forecast
European Commission Proposed Target: 50% Reduction Between 2000 and 2010
80000 70000 60000
Fatalities per year
1975
1980
1985
1990
1995
2000
2005
2010