How to interpret data

How to Interpret Data

By Charles Musing

Introduction

We hear numbers, figures, data, and dare I say "facts" every day. My research concludes that the overwhelming majority of national and social media intentionally or accidentally present these in a misleading manner, which is disheartening. It is counterproductive to honest conversation and solutioning for a better world, it is extremely divisive to people who have more in common than not, and it is damaging to the growth of human intellect. Media sources, especially large well-funded organizations, are in a position, given their resources, to be able to educate and inform, but largely they misinform and promote ignorance.

Purpose

The purpose of this article is to articulate basic principals in data interpretation and caution consumers, especially of media, to be very skeptical of facts, truth, conclusions and opinions being conveyed to them.

An example of how even real, factual data can be grossly misinterpreted is provided.

Storyline

The purpose of data can be broken down into the following categories:

• describe what happened
• explain why it happened
• predict what will happen
• prescribe what actions we should take

In a world where we consume messy, misleading information 30 seconds at a time (or less), we're all very rash to accept a truth, determine causation, and assign blame.

The below is an example of how scrutiny, patience, and most importantly, thoughtfulness are required to find truth and determine causality, and hopefully, not focus on blame but solutions and progress.

How information can mislead:

• it can be absent of concrete facts and completely or mostly opinion or feeling
  • we should be vigilant in deciphering the difference between thinly veiled opinion and indisputable fact
• it can be false
  • misinformation -- unintentional inaccuracies (errors, incompetence, or negligence)
  • disinformation -- intentional (lies)
  • This should be the easiest information to debunk and ignore. Unfortunately, it's effectives and often accepted at face value because:
    • it resonates with us and is consistent with our bias or ideology, or
    • it is so repetitious that it has marketing effectiveness, or
    • we lack the time, resources, and/or skills to disprove it
• it can be true, but critical additional facts are omitted
  • I find this to be the most dangerous of all deceptions
  • if enough facts are presented and those facts formulate a logical explanation to support the source's assertion, it can be very difficult for us to challenge it or fail to disprove it

Example

How_Statistics_Can_Mislead_--_A_Recent_Spike_in_Average_Wages_Caused_By_General_Increase_in_Unemployment

Key Terms

• correlation -- perhaps the most overused and misused statistical term. It indicates a mathematical relationship between two variables, from 1 to -1.
  • Why this is the most misused is that it is often used in conversation to simply mean one observed what appears to be an association between two variables, but an actual calculation was never performed. Without a calculation, there can be no correlation
  • While based on math, this measurement along does not imply or prove causality -- the change in variable b was caused by the change in variable a
  • 1 means that there is a perfect positive relationship: variable a increases and variable b also increase
  • -1 means there is a perfect negative or inverse relationship: variable a increase and variable b decreases
  • 0 means there is no correlation: an increase or decrease in variable a does not reliably coincide with either an increase or decrease in variable b
  • The Pearson correlation coefficient, r, is the most common correlation measure
  • I frequently prefer r-squared, the coefficient of determination

• association -- I dropped the hint above, this is the term that should be used in most conversations instead of correlation.
  • generic term with no mathematical calculation
  • simply based on observations where one variable's change appears to reliably coincide with another variables change
  • does not imply statistical significance
  • does not imply causality
    • Example: More soldiers are wearing military helmets, and more head injuries are being reported.
    • False conclusion, assuming causality: Wearing a military helmet causes head injuries; therefore, soldiers should not wear helmets.
    • Corrected conclusion: Wearing a helmet decreased combat deaths. The increased number of survivors of some head traumas are now reporting head injuries instead of going to the morgue with unreported injuries.

• causation -- the thing that everyone is trying to get right to: a change in variable a caused the change in variable b.

Further Reading

For those wishing to know more, I highly recommend resources like www.freakonomics.com. In particular, the books Freakonomics and Super Freakonomics have some fascinating insights and revelations of how interesting and dangerous data interpretation can be.

CharlesMusing (talk) 23:32, 26 November 2024 (EST)