5 Simple Lessons from Statistics to Drive Informed Decision-Making

“In any moment of decision, the best thing you can do is the right thing, the next best thing is the wrong thing, and the worst thing you can do is nothing.”
Theodore Roosevelt

Intro

Decision-making based on data is one of the most universally demanded skills of our time.

Every day, we face the decision to act, speak, or drive change in a particular direction. It can be as small as what you choose to eat for lunch, or as big as whether you choose to follow one career versus another.

What drives these decisions is often a mix of “gut instinct”, emotion,

#1: Using Descriptive vs. Inferential Statistics in Decision-Making

Descriptive statistics summarize data, while inferential statistics help make predictions or decisions based on data.

One common fallacy among non-statisticians is to confuse statistics that generalize about a population with statistics that have already been measured and established.

Imagine walking into a restaurant. You wonder what you will order.

If someone handed you the menu, you would see a clear outline of everything available. This is like descriptive statistics, giving you an exact picture of the sample population.

If you looked around, without a menu, and saw some people eating salads, you might infer that the restaurant orders salad. That is like inferential statistics.

But why does it matter to you?

We are exposed to statistical claims every day, both in professional settings and personal life. How we understand these claims directly affects how we respond to them.

If you want to respond better, start with your interpretation. Simple as that.

#2: Beware of Sampling Bias

Ensure that your sample accurately represents the population to avoid skewed results.

Sampling bias is one type of selection bias. Often, it occurs when a statistical study selects its subjects in a non-random way.

Importantly, this is not necessarily due to negligence. It can also take place whenever there are limited subjects available.

For example, imagine you are in that restaurant we mentioned, and you wanted to know what other customers recommend you order.

If you are only sampling customers in the restaurant right now, your responses will be biased. What if the usual chef was sick that day and the replacement messed up their orders? What if the lunch menu is terrible, but the breakfast menu is incredible?

If you only sample other customers who are physically in the restaurant with you at the same time, you will not get an accurate understanding of the best item to order.

Before jumping to conclusions based on other data in your life, consider this: you curate your own sample population; we all do.

What perspectives are you missing and how do you fill that gap?

#3: Understand P-values

A low p-value indicates strong evidence against the null hypothesis, but it doesn’t prove the alternative hypothesis.

First things first. What on earth are “p-values”, “null hypotheses”, and “alternative hypotheses”?

Essentially, a p-value tells you how likely it is for an event to come about as a coincidence. Low p-values suggest that the occurrence is unlikely to occur by chance, and vice versa.

When evaluating data, a null hypothesis is the starting assumption that there is no causal effect in what you’re studying.

The alternate hypothesis is the opposite of the null hypothesis and suggests a causal relationship in what you are studying.

The goal of an experiment is usually to either provide evidence against the null hypothesis or fail to find enough evidence to reject it, thus supporting the alternate hypothesis.

This is important for everyone to understand, as it highlights the important fact that not all results from data suggest causation.

Only the statistically significant results.

Speaking of causation…

#4: Decision-Making Based On Correlation vs. Causation

Correlation shows a relationship between two variables, but it does not imply causation.

The human brain is conditioned to recognize patterns, and in many ways this is thanks to its causal structure.

This is to say, we look for causation, not correlation, so it will be a perpetual struggle to untangle these from each other.

Nevertheless, we must try.

Identifying when correlation and causation are each at play is the key to thinking clearly about data-driven problem-solving, and solution analysis.

Consider this Washington Post article, which explores the correlation between crime and police spending.

Ultimately, the article concludes that there is “no correlation nationally between spending and crime rates”. However, causal research has shown that having more police typically leads to a reduction in crime.

See how easy it is to make this mistake, and the difference it makes?

#5: Beware of Confirmation Bias In Your Decision-Making

Avoid interpreting data in a way that confirms your preconceptions.

One quote commonly attributed to Stephen Covey is that “we see the world as we are”.

When using any data to make conclusions, it is vital to remember this.

The fact of the matter is this: statistics are not untouched by human psychology.

Actually, the human lens is exactly what gives any set of data true meaning. Because of that, it’s usually not “What is the data telling us?”, but rather “What do we want to hear?” The danger comes when what we want to hear from the data is aligned with what it is telling us.

Identifying this bias is important, not just for our self-awareness as individuals, but as scientists, policy-makers, journalists, business owners, and consumers.

By uncovering our own confirmation bias, we will be more open to identifying new patterns and developing creative new approaches.

Beyond this, addressing confirmation bias also serves to promote open-mindedness, which is crucial in a world of decision-makers faced with high levels of polarization.

To Sum It Up…

Understanding statistics is no longer a task just for researchers or analysts—it’s essential for anyone looking to make informed decisions, whether in business, healthcare, education, or everyday life.

By mastering these five concepts, you can approach data more critically, ask better questions, and ultimately make decisions that are not only data-backed but also rooted in accuracy and fairness.

Thought to Action

Diversity: Create diversity in any sample, survey, or research you conduct. Additionally, make sure to check for diversity behind the research you review as an outsider.
Dive Deeper: Instead of assuming a specific cause for an event, ask “What else could be going on here?”
Challenge Yourself: Approach claims with a beginner’s mindset, considering what are the gaps in your understanding of the data.
Beware of Data Dredging: Avoid searching for patterns in data without a prior hypothesis, as this increases the risk of finding spurious correlations.
Continuously Improve Data Literacy: Stay updated with statistical methods and best practices to make better-informed decisions.