Eight Things You Should Know About Statistics! according to
1. Assumptions: What are the assumptions underpinning the research? As seen from recent debate on CBO numbers for the U.S. health reform package, assumptions matter tremendously.
2. History: How much historical data was used in the study? What was the time scale? As seen from the 2008 financial crisis, the models used by Wall Street mavens often only took into account 10 years of data in judging the volatility and probability of failure of complex financial instruments.
3. Samples: Are the samples selected randomly? From what populations? Is there enough data for statistical significance?
4. Data Quality: The output is only going to be as good as the quality of data feeding the analysis. Garbage in, garbage out.
5. Survivorship Bias: Author Nassim Taleb points out that “losers are often not in the sample.” Does the analysis include a population of survivors and those who also failed?
6. Falsification and Omission: Yes, in an era of IPCC’s Climate Gate, one needs to ascertain if data are hidden, missing or outliers ignored.
7. Association equals causation fallacy: Correlation does not equate to causation (a common mistake made by marketing and finance executives alike).
8. Proper Application of Statistics: The effective use of statistics by insurance actuaries, scientists, and even casino managers is well-documented. However, real danger results when mathematical concepts are used to denote certainty indecision-making and divining behavior of markets.
1. Assumptions: What are the assumptions underpinning the research? As seen from recent debate on CBO numbers for the U.S. health reform package, assumptions matter tremendously.
2. History: How much historical data was used in the study? What was the time scale? As seen from the 2008 financial crisis, the models used by Wall Street mavens often only took into account 10 years of data in judging the volatility and probability of failure of complex financial instruments.
3. Samples: Are the samples selected randomly? From what populations? Is there enough data for statistical significance?
4. Data Quality: The output is only going to be as good as the quality of data feeding the analysis. Garbage in, garbage out.
5. Survivorship Bias: Author Nassim Taleb points out that “losers are often not in the sample.” Does the analysis include a population of survivors and those who also failed?
6. Falsification and Omission: Yes, in an era of IPCC’s Climate Gate, one needs to ascertain if data are hidden, missing or outliers ignored.
7. Association equals causation fallacy: Correlation does not equate to causation (a common mistake made by marketing and finance executives alike).
8. Proper Application of Statistics: The effective use of statistics by insurance actuaries, scientists, and even casino managers is well-documented. However, real danger results when mathematical concepts are used to denote certainty indecision-making and divining behavior of markets.
