HBX Final Reflections: Business Analytics — Cutting Through the Statistics Bullshit (#BusinessyBrunette HBX Week 9)
Welcome to #BusinessyBrunette! I am currently studying Harvard Business School’s HBX CORe, their online pre-MBA program teaching the fundamentals of financial accounting, business analytics, and economics. Every week I‘ll write up what I’ve learned — making it meaningful & accessible for artists, activists, geeks, nerds, fans, and anyone else who doesn’t fit the MBA Mould. I’m learning as I go, so feel free to critique, comment, tell me if I’ve messed up or did well, mash up, and share! [See the rest of the series here]
This is the final month of HBX CORe, so all the subjects are on their last modules. So from this week onwards I’ll be writing short reflections on each subject as they close out.
Business Analytics: Final Reflection
“Analytics” can sound very intimidating at first. You have to “crunch numbers” and “process data” and deal with terms like “regression” and “null hypothesis”. It’s basically applied Statistics, but “Statistics” often gets attached to sentiments like “urggghhhhhh I hate this Math class whyyyyy”.
I never really understood the hate against Statistics, probably because I hadn’t learned as much of it in high school as most people had. The short chapters we had that were Stats-related, I actually enjoyed, and would have liked to learn more. Most of my Stats education actually came late last year when I studied for the GRE — and this course was, in a lot of ways, a Stats crash course.
The terms are scary (I remember audibly groaning at “multiple variable linear regression”) but as I worked through the course, I found that a lot of them were actually very straightforward. You find out how likely it is that Some Result will happen, given your data. You think about what the status quo is, and then you figure out how it can be different and how to know when things have really changed. You take old data and use that to figure out how the future may look like. The myriad Excel formulae take care of the tricky math part — all you need to do is know how to read the results.
And knowing how to read the results leads to the biggest thing I’ve taken away from this course: how to cut through the bullshit.
It’s super common for people to misunderstand statistics — and for others to exploit that misunderstanding to further their agenda or cause harm: for instance, this chart about Planned Parenthood that ignored the Y-axis, or this chart about gun deaths in Florida that reversed the Y-axis to depict the opposite of reality. A basic understanding of how stats are compiled can help protect us against misinformation (intentional or meant well) and get us closer to the (often messier and less certain) truth.
Here’s some basic stats information for you:
- When something is “skewed” to any particular direction, that does not mean that most of the data is in that direction! It’s the opposite — a chart is skewed towards the outliers, the data points that are significantly different to the norm. So if someone says a chart is “skewed left”, it means that most of the data points are to the right of the mean (the average).
- Outliers aren’t necessarily irrelevant data points, nor should they just be omitted solely because they’re outliers. Rather, think about why the outliers were there: were all the data points collected the same way? Under similar circumstances? Do they answer the same question? If the data point is valid after all that questioning, then it’s worth finding out why those outliers exist.
- Subtle changes in the phrasing of survey questions can make a huge difference in the results, and leading questions can lead to really misleading responses. Similarly, the target samples and the askers for any survey can greatly affect the responses obtained, such as this claim that “most Britons want the UK to leave the EU” when the survey only ever asked readers of an anti-EU rag. What were the survey-takers being asked, who were being asked these questions, and who were asking these questions?
- You may see an R-squared number that seems high and thus looks like “oh there’s a strong correlation!”. However, that way lies spurious correlations. Not only does correlation not equal causation, there may also be hidden variables that aren’t being accounted for, and there are other metrics that need to be considered, such as the p-value. What is the study actually measuring? What data leads them to that conclusion? What are they not accounting for, and where are they drawing connections that don’t need to be there?
Statistics just sounds scary, but build up some basic knowledge and, at least, you’ll know enough to tell if someone’s trying to pull the wool over your eyes. (Or maybe you’ll know enough to do the wool-pulling. Your call.)
Other Business Analytics posts in this series:
Enjoyed this? Please support my work by subscribing to my Patreon and getting early access to my #BusinessyBrunette posts as well as other creative work! You can also make a one-off contribution or signal boost my work on social media! Thank you so much!
