Presentation on theme: "Foresight, Insight. Hindsight US Statistical Observations Fritz Scheuren NORC University of Chicago."— Presentation transcript:
Foresight, Insight. Hindsight US Statistical Observations Fritz Scheuren NORC University of Chicago
Reminder on Definitions Hindsight reflecting on the past –Personally/Collectively Insight, where it all comes together, like this Conference Foresight future seeing or shaping, also familiar but bears discussion
A Small Statistical Corner US Official Statistics Censuses and Surveys Administrative Records Focus on lived experiences, ala Deming
Times They are a Changing Relevance of our Discipline? Responsiveness of Statistics? Information Age? Misinformation Age? Service Partnerships?
Childrens Teaching Game? High? Low? Youre Too Slow! How Avoid Being Too Slow?
But Change is Accelerating! Is our Discipline Keeping Up? Certainly Computationally! Tool/Theory Building too! Practice Slower? Organizational Issues?
How To Keep Up/Catch Up? Google (of course) Metadata Revolution -- Still more Promise than Practice Meta-Analytic Reuse -- Still Often Too Hard
High-Clockspeed Trend ( use of cell phones, portable devices)
Responsiveness to Trends How Well Do We Play? High? Low? Youre Too Slow!
Response Times to Trends (organizational clockspeed = rate at which an organization introduces new products, adopts new production processes, or reorganizes itself; Sources: Charles H. Fine, 1998; David W. Rejeski, 2003)
Technological Mega-Trends Faster and Faster Computing (Slower for Official Statistics) Descriptive to Analytic Randomization-based to Model-based Producer Dominated to Customer Shared
Typical Grief Response to Change Still Often True Denial Anger Bargaining Depression Acceptance
Examples of Hindsight, Insight, Foresight Nonresponse Circa 1980 US Census Taking Circa 1990 Paradata Circa 2000 Visualization Circa 2010 Next Steps Together?
Nonresponse Hindsight Example
40 Year CPS Income Trend Insignificant Missingness in 1962 Now Nearly Half of Interviews have Some Missingness About a Third of the Amount is Imputed But Still using the Same Basic Hot Deck Methods Today
Greater Bias Concerns Possibility of Greater Nonresponse Bias Potentially More Income Understatement Also Characteristics of Poor Blurred
Variance More Understated Growing Variance Not Directly Reduced Rubins Multiple Imputation Solution Still Not Used in CPS Remains Descriptive Rather than Analytic
Paradata Modeling Insight Example
Meta-Data Revolution Applying Computing to Documentation and Training Including Measurement Process or Paradata Achieving Full Systems Thinking
Unify Meta/Paradata Bringing All Survey Aspects together electronically Sharing with All Stakeholders Breaking Down Barriers between Departments
Unify Meta/Paradata Bringing All Survey Meta- Data together electronically Sharing with All Stakeholders Breaking Down All Barriers Between Departments
Manage System as a Whole Not Just Conformance to Requirements Quality But Total Fitness for Use Quality From Sampling/Nonsampling to Total Survey Inference Record linkage Example
Total Systems Thinking Turning Sample Models Into Full Survey Models Using Paradata and Experience Politz-Simmons Example
Visualization Foresight Example
Complex Survey Graphics Clustering and Weighting Distort Analytically these can be solved Approximately Design Effect Example
Restoring Visual Metaphor Inverse Sampling Algorithm Works for Many Designs Satisfactory Analytically Works Graphically too but not yet Always