Statistician Mark Hodge and a team of researchers at Johns Hopkins University in Baltimore recently released a report examining the different definitions of what constitutes “statistic evidence.”
Hodge wrote in the paper, “We hope that this report will help inform discussions about how to define and classify statistical evidence in the public interest.”
The researchers define statistical evidence as “evidence that predicts, predicts, or provides a reliable explanation for an observed phenomenon, event, or trend, based on its statistical significance.”
“A causal relationship exists between an observed event and an outcome or state of affairs that is not fully explained by a random, non-causal process,” the report reads.
The definition of “evidence” can also be subjective, as the researchers noted.
Some researchers are using the term “statistical evidence” to refer to the statistical tests used to find a relationship between a given event and a result.
“While a causal relationship is evident in the statistical test used to predict an outcome, it may not be the case if a particular statistical test or its results are used in conjunction with a causal theory,” the authors write.
“In fact, if a causal test is used in a way that is incompatible with a theoretical model, the test may not predict the outcome at all.”
Hinkle, who is also the founder and executive director of the Center for Data Analysis and Data Science at Johns Hopkins, said that he and his team are working on “improving” the statistical evidence definition and are working with other experts to come up with a more consistent definition.
“I think we’ve identified the problem, but we’re not done yet,” he told Business Insider.
“We’re not there yet.”
A similar problem is also apparent with the word “statistics.”
Some researchers define the word as “the systematic study of the nature, properties, and processes of statistics.”
Others define the term as “a method of applying statistical knowledge to solve a problem.”
But when you use the word statistical, you are only defining what you believe the word means.
For example, statistician Richard Lynn, who coined the term, has said that statistical evidence does not define the data.
“Statistical evidence is about making sense of data,” Lynn wrote in a statement on Twitter.
“It is not about making a hypothesis.”
But, Lynn added, “statists can still make sense of the data using statistical methods.”
Lynn is not the only statistician who has called the word mathematical evidence.
For decades, statisticians have used the word to refer both to the mathematical process that determines what the data is, and the statistical procedure that is used to test the data and the conclusions drawn from them.
“There is nothing mathematical about it,” said University of California at Davis statistician James Buechler, who has published dozens of books and many papers on statistical methods.
“The thing that makes mathematical data interesting is that there is a relationship to the underlying mathematics.
And that is how people interpret the data.”
For example: Lynn wrote about how mathematicians can use statistical methods to study patterns of poverty in the U.S. “To use mathematical techniques to investigate poverty in America, it is useful to be able to look at data and use the techniques that are relevant to understanding poverty in a particular community,” Lynn said in a 2016 interview with the Guardian.
“A mathematical analysis can give you a sense of what might have happened if the poverty rate had not been so high.”
Statistical evidence is often used as a way to prove that certain hypotheses are correct.
In the past, statisticists have used statistical methods that were considered controversial, but that were eventually adopted by mainstream economists.
For instance, researchers have used regression to estimate the effect of tax cuts and other policy changes on employment.
However, the term statistical evidence can also apply to other areas of study, such as the science of statistics, the use of mathematical tools in the development of medical diagnostics, and many other areas.
A new definition of statistical evidence is needed to make the term more consistent with the broad definitions of statistical and non-statistical methods.
But, the new definition should be considered a first step in making the word more widely used in the scientific community.
“These terms have always been problematic and there has always been a debate about what statistical evidence really means,” Lynn told Business Insider.
“But in this day and age, people can use these terms to refer not to a specific method or method of testing, but to a broad category of tools that can be used to do analysis.”