It’s easy to think that if you just knew statistics better, data analysis wouldn’t be so hard.
It’s true that more statistical knowledge is always helpful. But I’ve found that statistical knowledge is only part of the story.
Another key part is developing data analysis skills. These skills apply to all analyses. It doesn’t matter which statistical method or software you’re using. So even if you never need any statistical analysis harder than a t-test, developing these skills will make your job easier.
Statistical knowledge is necessary to even get started developing these skills. And as you develop the skills, you’ll find the statistics make more sense. But more statistical knowledge isn’t a substitute. So what are these important skills?
1. Planning the Data Analysis
Like most projects, data analysis projects are more efficient and have fewer issues if you have a plan. A plan requires you to think ahead on critical decisions that will be time-consuming to redo.
The correct statistical analysis depends on the research question, the study design, the variables, and any data issues. Considering how these work together before you start can save a lot of time and headaches during the analysis.
2. Managing the Data Analysis Project
Even if you are the only person running the data analysis you still have to manage the project itself. You need to keep track of files, allot enough time to each step, and find the resources you’ll need along the way.
3. Cleaning, coding, formatting, and structuring data
You’ve heard the term GIGO, right? Garbage In, Garbage Out. In data analysis, it’s the data that is going in. And it needs to be clean. Sparkling.
And once it’s clean, you need to code and format the variables. Then you need to structure it correctly for the planned data analysis.
This step often takes much, much longer and requires more statistical software skills than the data analysis itself.
4. Running analyses in an efficient order
There is a specific order in which to run the steps of your analysis and there are decisions to make at every step.
If you do them out of order or hit a roadblock you can’t solve, the analysis will be slower and more frustrating. More importantly, you’re likely to make mistakes.
5. Checking assumptions & dealing with violations
Yes, every statistical test and model has its own assumptions. So the content of this skill will differ depending on what you’re running. But the general approach to checking assumptions is the same. And there is a lot of skill in reading uncertain situations and drawing conclusions.
6. Recognizing and dealing with data issues
Real data are messy data.
Real data have issues that make the analysis hard. Outliers, small sample sizes, and truncated distributions happen in all types of data sets. Recognizing when a data issue is happening, knowing whether it’s bad enough to cause problems, and knowing what do to about it are important skills.
7. Detecting problematic results and troubleshooting
Sometimes you get weird results — really weird results — despite cleaning data, checking assumptions, and looking for data issues.
There are so many possible causes: typos in the data set, glitches in the software, missing data.
The skill here is being able to recognize when something is wrong and how to investigate the issue and the solutions.
8. Interpreting, Presenting, and Communicating Results
This set of skills may be the most important of all. It includes interpreting results and writing them so the audience understands. It requires creating useful, appropriate, and accurate graphs and tables.
This also means knowing how to get your stat software to do the heavy lifting for you. If you can create the tables the way you need, you don’t have to spend hours reformatting them.
How to develop these skills
No data analyst starts out with these skills, no matter how many statistics classes they’ve taken. There is only one way to develop these skills: it takes experience analyzing real data sets.
But the skill development comes easier if you have specific training on these skills and someone to mentor you as you gain experience. Think of yourself as an apprentice of a skilled trade. Yes, you can do it on your own. But you will improve your skills faster and produce better work along the way with some guidance.
Good list. I would put #3 at the top, but I agree that the other items are also important.
Thank you so much Karen, this is helpful.
I am interested to join and learn much about statistical analysis if given an opportunity. Many thanks and more power to your institution.
Thanks very much for simplifying presentation to enhance understanding. You are highly appreciated.
Thanks, Oyeleye!