Communicating conclusions

Think of 'communicating conclusions' as telling a really good story with numbers. You've been investigating a question, like 'Does more sleep make you score better on tests?' You've collected data (how much sleep people got, their test scores), done some fancy math (regression analysis), and now you have some answers.

Your job is to explain those answers in a way that anyone can understand, even if they've never heard of a p-value (a number that helps us decide if our results are real or just by chance) or a confidence interval (a range of numbers where we're pretty sure the true answer lies). It's like translating a secret code into plain English so everyone can get the message. You're not just showing the numbers; you're explaining what they mean in the real world.

We focus on inference for regression (making educated guesses about a whole population based on a sample when we're looking at the relationship between two numerical things). So, you'll be explaining what the relationship between two variables, like study time and test scores, looks like for everyone, not just the people you studied.

Let's say a local ice cream shop wants to know if advertising more on social media actually leads to more ice cream sales. They track their weekly social media ad spending and their weekly ice cream sales for a few months. After collecting the data, they use regression analysis (a statistical tool to see how two things are related) and find a relationship.

Their statistician, you, then has to explain the findings to the shop owner. You wouldn't just say, 'The slope is 0.5 and the p-value is 0.01.' That's like speaking in code! Instead, you'd say something like:

'Based on our study, for every extra dollar we spend on social media advertising each week, we can expect to sell about 50 cents more in ice cream. We're pretty confident this isn't just a fluke, and we're 95% sure that the true increase in sales for every extra dollar spent is somewhere between 30 cents and 70 cents.'

See how that's much clearer? It tells the owner exactly what the numbers mean for their business decisions.

1. State the Question: Start by reminding everyone what you were trying to figure out. What was the main goal of your study?
2. Summarize Your Model: Briefly explain what kind of statistical model you used (like a linear regression model) and what it showed. Mention the slope (how much one thing changes when the other thing changes) and the y-intercept (where the line crosses the vertical axis, meaning the predicted value when the x-variable is zero).
3. Interpret the Slope in Context: Explain what the slope means using the actual names of the things you studied. For example, 'For every extra hour of study, test scores are predicted to increase by 5 points.'
4. Interpret the Y-Intercept (if appropriate): Explain what the y-intercept means in the real world. Sometimes it doesn't make sense (like predicting test scores for 0 hours of study), so mention that.
5. Discuss the Strength of the Relationship (R-squared): Explain how well your model fits the data using R-squared (a number that tells you what percentage of the changes in one variable can be explained by changes in the other). Is it a strong relationship or a weak one?
6. State Your Conclusion about the Relationship: Based on your p-value, decide if there's a statistically significant relationship. This means you're confident that the relationship you observed isn't just due to random chance. Use phrases like 'There is convincing evidence...' or 'There is not convincing evidence...'
7. Provide Your Confidence Interval (if applicable): Give the range of values where you're pretty sure the true slope (the real relationship in the whole population) lies. Explain what this range means.
8. Acknowledge Limitations: Briefly mention anything that might have affected your study or what you couldn't conclude. No study is perfect!

1. ❌ Using Jargon Without Explanation: Saying 'The p-value was 0.03, so we reject the null hypothesis' without explaining what a p-value is or what 'rejecting the null hypothesis' means to a non-statistician. ✅ How to Avoid: Always explain statistical terms in simple language. 'The p-value of 0.03 means there's only a 3% chance we'd see results this strong if there really was no relationship. Because this chance is small (less than 5%), we have strong evidence that there is a relationship.'
2. ❌ Confusing Correlation with Causation: Saying 'More social media ads cause more ice cream sales' just because you found a strong relationship. ✅ How to Avoid: Remember, a strong relationship (correlation) doesn't automatically mean one thing causes the other. It's like saying 'Ice cream sales go up when more people wear shorts.' Shorts don't cause ice cream sales; warm weather causes both! Use careful language like 'is associated with,' 'is related to,' or 'is predicted to increase with.' Only claim causation if your study was a well-designed experiment (where you control everything).
3. ❌ Extrapolating Too Far: Using your model to make predictions far outside the range of your original data. If you studied ad spending between $100 and $500, don't predict what happens if they spend $10,000. ✅ How to Avoid: Always remind your audience that your predictions are only reliable within the range of the data you collected. It's like trying to predict how tall a 50-year-old will be based on data from 5 to 10-year-olds; it just won't work!