Medium has always been a mystery of what works and what doesn’t work. Sometimes an article I expect to do well does not do well. Sometimes an article surprises me with more traction than I had expected.
If I were Pavlov’s dog and Medium were a laboratory doling out food rewards, I would probably starve.
The article explains my finding with placing articles under Medium publications, using a statistical approach.
The Student’s T-test
It is one thing to analyze data with a squint or a gawk and make a sound conclusion in this way. But professional statisticians will not accept our squints and gawks. To gain their acceptance, we need to run our numbers through a credible statistical process, then reach our conclusion.
I will use the “Student’s T-test.” This test compares two populations. Each population set has its own mean average and standard deviation. To give an example, I have hand-drawn four charts. Each chart has two populations. The population with higher mean is to the right. The population with more variance is wider. These curves are normal distributions.
Chart 1 shows two populations that are different. Chart 2 shows two populations that are similar. Chart 3 and 4 are not so conclusive. We would need the student’s T-test to determine whether these charts show similar or different populations.
To determine whether two populations are similar or different, we put the mean and deviation into the formula below:
From that formula, we can discern whether one population is indeed “better” than the other.
For this particular study, I applied a few more statistical parameters into the t-test. I won’t get into exact details, but I came up with a T-Value of 2.31. This means:
· If the T-score is greater than 2.31, that means there is at least a 95% probability that the two populations are different. Many researchers would base their conclusions that a difference exists between the two populations. This 95% confidence level is a traditional application.
· If the T-score is less than 2.31, that means there is not enough evidence to conclude that the two populations are different. Researchers would assume similarity in their conclusions.
· For T-scores closer to 0.00, that would signify a certain similarity.
In this study, we will be using the above formula and a T-score of 2.31 to determine if certain pubs are better than the others, in terms of Medium shekels coming my way.
My Current Pubs
Independent Publishing
Quite often, I have an article that does not fit my preferred pubs. Rather than trying to find a pub that could accept this article, I just publish independently.
This group constitutes my control group. Whatever traction this group gets is because of my following and whatever exposure Medium usually gives me, not because of a publication’s following.
Tiered Democratic Governance (TDG)
This is my own pub. I am the only writer. Sometimes my articles get pretty heavy with my alternative democracy. I park these heavy articles in this pub because it will easier for new TDG fans to find my more important work on Medium.
Unpopular Opinions
Well, my alternative democracy has proven to be unpopular. But I have other ideas that are also unpopular with the Medium echo chamber. Melanie J. has a pub for such unpopular opinions. I have been publishing articles on this pub for three years, often with a TDG slant.
Free Factor
For almost four years, I put my lighter political articles on the Medium pub called Politically Speaking. Owner Scott Tarlo and his two assistants volunteered a lot of time making articles a little better and fitting them into the Politically Speaking style guide. This extra work made Politically Speaking quite popular. But there was a price to pay. Scott burned out, and maybe his two helpers. Politically Speaking was suspended.
If the Medium gods are listening, I believe the Medium shekels need to be shared between the writers and pub owners, maybe an 80/20 ratio.
I wanted another pub to put my lower-TDG-content political writing to see if I could attract new readers. There are political pubs on Medium, but they seemed to attract the popular anti-Trump, anti-MAGA rhetoric. I didn’t see much point preaching to this audience.
Then I found Colby Hess has set up his own political pub called Free Factor. This pub is not another political echo chamber; it welcomes well written, yet off-the-beat political analyses. I have been publishing here for a year.
Political Prism
John Polonis has a similar approach with his pub, Political Prism. I am a recent new writer to this pub.
When I have a political article with little or no TDG content, I alternate between these two pubs. So which pub works better for me? You will have to read the rest of this article and see what the T-test tells you about these two pubs.
Hypothesis
Supposedly I have 2.2k followers. I believe many of them have left Medium, were fickle followers, or were bots. I estimate that 200 people see my articles on their feed.
If my Medium shekels come from mostly from my followers, then the five pubs should show that they are not that much different than each other. The T-scores will be close to zero.
If a pub’s followers bring new eyes to my work, we should see higher T-scores. In other words, a higher T-score show the pub has some effect on my traction (if the pub has a higher mean average).
My hypothesis is that there really is not much difference between the pubs.
Dave’s Data
I used the last five articles in each pub for this analysis, for a total of 25 articles. The table summarizes the statistics:
Then I put each pub head-to-head with the other pubs. Here is the T-score for each contest. The bigger the score, the more likely the pub with the higher mean is the better pub for Medium shekels coming my way.
Independent vs. Tiered Democratic Governance: 1.17
Independent vs. Unpopular Opinions: 1.46
Independent vs. Free Factor: 0.00
Independent vs. Political Prism: 0.16
Tiered Democratic Governance vs. Unpopular Opinions: 2.41
Tiered Democratic Governance vs. Free Factor: 1.19
Tiered Democratic Governance vs. Political Prism: 1.11
Unpopular Opinions vs. Fear Factor: 1.48
Unpopular Opinions vs. Political Prism: 1.73
Fear Factor vs. Political Prism: 0.17
Statistical Conclusion
According to conventional students’ T-Test applications, professional statisticians would claim that only two pubs showed a difference to each other. Unpopular Opinions was better than Tiered Democratic Governance, in terms of Medium shekels coming my way.
In general, the statisticians would say there is not much difference between the other pubs. In other words, other pubs were not instrumental to helping my Medium traction.
Dave’s Conclusion
I had a contention with the students’ T-test when using it 40 years ago. That contention is still there. It is the reliance on using the 95% confidence level.
In this study, there are six T-scores between 1.1 to 1.8. These values would signify a confidence level around 50% that the two populations are different. Yet because they did not reach the 2.31 value, the statisticians would claim similarity, not difference. But it is hard to believe that all six of these comparisons would side with similarity. So, in my amateur opinion, the full story is not told with the 95% confidence level.
I would rather turn a T-score into a probability of difference rather than a yes-no answer. But so much scientific and humanistic research has been done based on 95% confidence levels that it would be hard for statisticians to move beyond this tradition created many decades ago.
Now that I have just bashed professional statisticians for holding on to outdated techniques, I will give my analysis, based on my interpretation of the T-scores of the 10 competitions.
The T-scores show three distinct groups.
Unpopular Opinions is clearly at the top. This pub has been around for a long time and acquired many followers. I estimate that at least a third of my traction from this pub came from the pub itself.
Free Factor, Political Prism, and Independent are in the middle. This suggests that the two pubs are not generating much extra traction for me. But (I think) these two pubs have not been around as long as Unpopular Opinions and have fewer followers.
Tiered Democratic Governance is clearly at the bottom. While this pub has been around for four years, it has failed to find willing readers. Methinks when the acronym “TDG” shows up in a title, my lukewarm followers know that I’m going to admonish them for trying to bend the will of their favorite non-fascist political party instead of building the new democracy I am telling them to build.
I have a lonely life! Now it has been proven statistically.
Caveat: When I did my last student’s T-test 40 years ago, I had a statistical guru to guide me. This time I did not consult with a statistician. Maybe my application of this statistical tool is not correct.
Another Experiment
Democracy is a Greyscale is my worst article on Political Prism. It only got $0.10.
But notice the title: it had the words “is” and “grey.” Not words to entice readers to click.
Had I written a better title, I believe it could have earned $1.50. This result would have increased this pub’s mean and decreased its standard deviation. These changes would mean this pub faring better in the head-to-head competitions for T-scores. So I reworked the numbers, replacing the $0.10 with $1.50. The T-scores for Political Prism did improve, but nothing came close to 2.31.
And a bit more
I mentioned the suspended pub Politically Speaking, where I published at least 50 articles. Because this pub has been suspended for two years, I really should not compare its performance with the other pubs. However, I will say that with my squint-and-gawks recollections, Politically Speaking did perform a little better than my independent publishing and my own pub. But I doubt Politically Speaking would have reached the 95% confidence level.
And I should thank Melanie, Colby, and John for putting up my articles on their pubs. I know this task takes a little volunteer time.
Again, I hope Medium finds a way to give a little cut to pub owners. I think this would change the dynamics of Medium. More “mostly readers” would gravitate to the better pubs — and be willing to pay subscription fees. The pub owners need at least a token financial incentive to enhance their efforts to design pubs readers are attracted to.
What’s next?
I might do this analysis next March. It was kind of fun. It would be interesting to see if Free Factor and Political Prism are rising above my independent articles.
Further reading
If you are interested in applying the student’s t-test to your Medium data, here is the website I used. It was easier to use this website than my 40-year-old statistics textbook.
https://www.scribbr.com/statistics/t-test/
Published on Medium 2024
