Issue #36: How I write my scripts

🚨 NEW VIDEO DROP

The coding video was dedicated to teaching how to use a proportional hazards model in R.

📺 What’s coming up?

Code for Bayesian linear regression; maybe a Monty Hall thing?

📰 TL;DR

Outlining is 80% of the process.

In this issue…

I wanted to document how I approach doing research for videos. I anticipate that this will change in the future, so I’m documenting it for posterity.

When I start writing a video, I always start with an outline. This outline helps me layout the main topics I want to talk about for a video and it gives me a built-in method for reviewing my writing multiple times before it gets brought into voiceover. I’m the only one who works on the channel, so every bit of feedback I can give myself helps.

I used to think of writing an outline as lesser task than actual script writing, but it’s proven to be so useful that I never do a video without one.

The outline is essentially just a bullet point list containing the sequence of points I want to cover. The first things I write are the main sections of the video. For example, with the most recent video on the t-test:

1. Hook: have you ever wondered why the t-test is everywhere

2. Background and some history

3. General hypothesis test strategies

4. Likelihood ratio

5. Exponential families

After all the main beats written, I start to think about all the points I need to make to fill in between these sections. This includes basic concepts that need to be defined, questions I want to make sure I ask the viewer, key equations, etc. I’ve found that it saves time to try to write these bullet points in the same way I would say it out loud. I tend to write long sentences, so this helps stop this tendency.

Crucially, forming this bridge between the main sections helps me identify gaps in my knowledge. As a first pass, I try to write down as much of the outline based on information in my own head. This just saves time. Inevitably, there are details I forget or stuff I know but am not sure is true or needs proof. I try not to switch tasks when I’m outlining so, I often write unknown information as another bullet point to myself.

For example:

4. Likelihood ratio

   1. There are theorems that make UMP tests…

   2. But does Karlin-Rubin apply to the t-test?

Once an initial outline is done, I start to research old lecture notes, textbooks, research papers, and online resources. I always hope that my lectures or textbooks have the answer I need because verifying stuff online compels me to look for multiple sources that say the same thing. That being said, I do enjoy the opportunity to look at the historical papers that past statisticians have written. The history of the statistics and its development does not have the same awe-inspiring feel as does math.

I bring in this extra research into the outline and then start with script writing. The hook is — by far — the hardest part to write because I find it’s inherently hard to convince a stranger that statistics is interesting. I often alternate between writing the middle of the script and the hook so I don’t waste time being hung up. I used to not care that much about thumbnails and titles, but in my goal to reach 100K before I graduate, I feel compelled to play the YouTube game a bit more. I am trying to be more thoughtful about my thumbnails and hooks since it’s really true that no one will watch my videos if these don’t convince people to stop and click.

The middle of the script is easier. If I’ve done the outlining right, then I can just put my outline in a paragraph. It doesn’t always end up like that, but it usually does.

A written script will only go into voiceover after it has been polished at least once. By that, I mean that I’ve given myself a day away from a part that I’ve written and have gone back to check. If I’m in a time crunch, I will often write and polish in sections so that voiceover can be brought into Final Cut for me to edit. I get sick of writing for long periods of time, so being able to edit something in the meantime keeps me productive.

A part of me wants to be able to work from the bullet points though in the form of talking on camera. This would greatly save time on finding stock footage, but brings extra complications. Whatever makes video making easier because it’s always a big mental lift.

Thanks for reading, see you in the next one.

---

Footnotes

🧐 What am I enjoying right now?

• Severance just finished, so life feels a little empty right now.

• I’m still listening to The Minimalist Entrepeneur by Sahil Lavingia. It’s been a challenge to figure out how to make something that can be useful to my viewers, but it’s been a fruitful experience to think of ideas.

📦 Other stuff of mine

• I wrote guided solutions to problems to Andrew Gelman’s Bayesian Data Analysis. It’s for advanced self-learners teaching themselves Bayesian statistics

• You can support me on Ko-fi! YouTube and Substack are by far the best (and easiest) ways to support me, but if you feel like going the extra mile, this would be the place. It is always appreciated!