240 likes | 372 Views
From Surveys to Surveillance Time Series Analysis. Under Construction !. Eric Holowaty, Sr. Scientist, Informatics Unit, CCO Michael Spinks, Sr. Res. Assoc., CCO. Background. 1980s and 1990s - occasional surveys precise estimates of rates, proportions, means, nos. affected
E N D
From Surveys to SurveillanceTime Series Analysis Under Construction ! Eric Holowaty, Sr. Scientist, Informatics Unit, CCO Michael Spinks, Sr. Res. Assoc., CCO.
Background • 1980s and 1990s - occasional surveys • precise estimates of rates, proportions, means, nos. affected • detecting important differences in estimates within survey • 2000s - surveillance systems • continuous, or at least frequent sampling • monitoring and assessing temporal patterns, incl. change point detection and sub-group differences over time
Background • What is desired periodicity of sampling? Depends on: • how rapidly variables actually change • how important it is to detect changes quickly • desired precision in describing temporal patterns, changes and differences
Definitions • Time series • sequence of data points, measured at successive times, and spaced apart at uniform time intervals • Time series analysis • methods and models that describe and explain temporal patterns, and forecast future patterns • Trend • long-term movement in an ordered series; may be temporal or just ordered strata
Risk Factor Surveillance in Ontariopre-RRFSS • Uncoordinated • Fragmented • Lack of smaller area data • Poorly analysed • Poor dissemination • Not timely • Difficult to access
Pilot tested in Durham Region in 1999 • Available for Individual PHUs in Jan 2001 • 22 PHUs participating as of Dec 2004 • ?Province-wide coverage in 2005/06
RRFSS Population Coverage 87% of pop’n 22/37 PHUs RRFSS (2003) respondents : 25,600 CCHS (2003) resp. : 37,000
Benefits of RRFSS • Monthly data more suitable for detecting temporal changes • More flexibility re. aggregation - before / after comparisons; geographic areas; demographic groups • Seasonal effects can be better described and analysed • (Robust SPC procedures permit timely detection of stat. signif. changes) • LARGE sample size permits more precise analysis • Standard CORE of questions helps ensure comparability over time and with other geo. areas. • Flexible MODULES permit targetted sampling and invest. of local concerns
Fundamental Statistical Issues in Time Series Analysis • Accuracy and precision of estimates • precision ~ sample size and survey design • bias • differential access and response • reporting/measurement bias • changes in the measurement tool, incl. wording importance of bias in time series analysis depends on size and consistency • Statistical power • probability of detecting an important change in time series - slope; seasonality; change points
Estimating the rate of change over time • Estimating that slope differs from the null i.e., zero change • assumption of monotonic relationship e.g., linear or log-linear model or logistic • assumption of no change points in time series
Statistical Power to Detect Slope > Null • Power influenced by: • length of time series (k) • size of each sample (n) • measurement of interest (p or x or x) and its variance • alpha (Type I error) • underlying rate of change/slope (b)
Statistical Power of Trend Tests Sample Size From: MacNeill and Umphrey, 1997.
Statistical Power of Trend Tests Sample Size From: MacNeill and Umphrey, 1997.
Statistical Power of Trend Tests Sample Size From: MacNeill and Umphrey, 1997.
Monthly Estimates of ETS Exposure Trends - RRFSS GTA Aug01-Dec03
Quarterly Estimates of ETS Exposure Trends - RRFSS GTA Aug01-Dec03
Detecting abrupt changes in lengthy time series • Change-point methods e.g. JoinPoint • Control Charts • conventional p-charts • CUSUM charts • EWMA charts, with residuals
Monthly Estimates of Support for Bylaws - RRFSS GTA Jan02-Dec04
Monthly Estimates of Support for Bylaws - RRFSS GTA Jan02-Dec04 15/35 point estimates in violation of Western Electric rules
Monthly Estimates of Support for Bylaws - RRFSS GTA Jan02-Dec04
Monthly Estimates of Support for Bylaws - RRFSS GTA Jan02-Dec04
Plan • Complete analysis of definitions, incl. temporal consistency and CCHS consistency • Assign final sample weights • Production of point estimates for 2003 and 2004 • Age-standardized comparisons • Time series analysis