Some students avoid statistics. Others dive in headfirst. One student I met this week did the latter—spending hundreds of hours teaching himself statistical analysis from scratch using textbooks, Google, and R Studio. And no, he did not use AI.
By the time he came to me to check his results, he had done a solid job. His analyses were mostly correct, but there were gaps—he hadn’t computed effect sizes and had missed some additional tests. The problem? He didn’t know what he didn’t know. But his hard work paid off—he understood his results and could write an excellent discussion chapter. Kudos to him.
This student was an outlier. Most students I meet are happy to hand their data over to a statistician in a heartbeat. But statistics is one of those skills that you can pick up on the go—especially now, with AI tools like ChatGPT, Julius AI, and others. The challenge? These tools will give you results, but unless you know what analyses and assumptions to check, you might not realize when something is off.
Curious, I tested this myself. I took a dataset I had analyzed conventionally (using Stata, etc.) and then reanalyzed it using AI, playing the role of a student with little statistical knowledge. I described the data but didn’t mention its skewness—just as a beginner might overlook this step. ChatGPT guided me on what to look for, and Julius AI ran the tests. The result? Not flawless. It performed parametric tests on highly skewed data, producing questionable results.
The takeaway? AI can be a helpful guide, but you must stay one step ahead. If you don’t know what to check, you won’t spot mistakes until it is too late.
So, get involved in your own data analysis. Play with your data. Use AI as an assistant, not as a leader. Most importantly, you should always get your statistics checked.
Have you used AI for your own data analysis? What was your experience?
Contact me at [email protected] if you need further information or help with your statistical analysis or any aspect of your dissertation.