My favorite practices to help you appear a Smart and Confident Data Science professional.
Appearing smart should be the top priority of every Data Scientist. This is what drives your career, right? Sarah Cooper's advice on how to appear smart in meetings is a fantastic start, but to succeed among Data Scientists is a whole other ballpark. We're talking about impressing a room chock full of Ph.D.'s and self-educated whizzkids. But fear not, with a few simple tricks you can deceive any human.
1. Draw a bell curve
Standing up and drawing things is great for your personal exposure.
Normal distributions can be found in all data somewhere, so there is no better way to show off your statistical skills than by drawing bell curves. On top of that, drawing one is bound to ignite a discussion about other assumptions there may be about the data in question. After which, as Cooper states, you can go back to playing with your smartphone.
2. Ask about the outlier
Data Science meetings are about data and data visualizations presented in some form. And there is always something wrong with the data. So when there is a silence during the meeting, ask: "Can we go back to that slide with the outlier?". This is a guaranteed success!
Note: should you find yourself on the presenting end, always make sure you leave a few outliers in your figures that you can easily explain. Works like a charm.
3. Nod in agreement when you hear about overfitting
When you are dozing off during a boring discussion and you hear the word overfitting, make sure you appear alert and nod aggressively in the direction of the speaker. Overfitting is the root of all evil! All experienced Data Scientists have learned to fear it. Acknowledging this fact not only grants you instant intelligence but when you meet again they will remember you were at their sides in the face of a common enemy.
Should you be in the occasional odd meeting where someone asks "What is overfitting?", simply redirect the question to the most junior engineer for clarification (who most certainly took a Data Science course recently). This makes you look more like a leader.
4. Ask about the business value
In the end, the business value of your analysis or product is the only thing that matters, right? But Data Scientists just want to solve fun puzzles, make cool graphs and appear smart. So this will get them running in crazy mental circles!
5. Suggest a random algorithm
Just memorize this list of algorithms, and randomly shout one that hasn't been mentioned yet. You gotta try em all.
Linear regression (always a classic)
Lasso regression
Random forest
Anything with the word Bayesian
Neural Network (or quickly google the latest hyped one, for example, Capsule Networks)
Principal Component Analysis
k-Nearest Neighbours
Feel free to expand this list as you gain experience in buzz-words mentioned in Data Science meetings and online articles.
6. Skew the used metric
When the someone mentions their method has a 10% error rate, immediately take over and proclaim that means we have found a 90% success rate! It is also great to change to fractions of fractions. Going from 90% to 95% success rate means a 50% reduction in the error rate! People will feel great about minor accomplishments and they like you even more for pointing it out.
If you do not understand a figure, it is best to turn this into an asset. But don't forget to put the blame elsewhere. Suggest that the managers or customers of your company may not understand it this way. The graph will be much improved if the presenter just "Keeps it Simple".
Also, say "Less is More"at random intervals and refer to Edward Tufte. No one ever really read his book, but this makes you look like a true aficionado.
8. Make a nerdy joke
Funny people are perceived as more intelligent. Funny nerds must, therefore, be on top of the smartness hierarchy. What could possibly go wrong?
9. Organize a Data Hackathon
Best. Meeting. Ever. You invite all the smart people to your meeting room. Give them your data and problems. And then they solve them for you, with nice scripts and figures. Finally, you give them points based on their contributions. In the end, they will even thank you for everything!
10. Ask deep questions, like "Why?"
Typical Data Science project flow
The worst thing that can happen is that your project ends and you have to start all over impressing new people on new projects.
So how do you keep the wheel going? With the well-known data science flow for projects, you can make it explicit how to do this. The best approach is to force everyone to go back to the beginning. Ask things like or "Why were we doing this again?" or just "Why?". It's called Data Science for nothing; science is all about asking neverending questions!
Matthijs Cox is a high-tech professional from The Netherlands, specializing in Nanotechnology and Data Science. He likes to talk about himself in the third person, play with his children's Lego, create beautiful images and design algorithms that may take over the world. He's only kidding about one of the above.
