Hacker News Reader

Comment by one-punch | Hacker News Reader

> If so then we should look at the slope of the graph for 'growth', and it in fact appears to have flattened recently.

The second graph shows the number of job posts per year, and it seems to reflect the bigger macro-economic situation: dip during pandemic, and a slight drop after 2022. I think these features are not specific to Haskell, just showing that the Haskell segment also follows some bigger trend.

View on HN

If the second chart is accurate then the first chart should be showing continual upward progress over time, but what we see is that the first chart is starting to flatten out around 2023.

Either the term "cumulative" is being used incorrectly, or different rules are being used to classify available jobs in each graph.

Maybe the first chart is filtering out duplicates?

The second chart is bucketed by year, while the first chart is not (appears to have more continual changes). Also, the discrepancy you observed around 2023 is on the edge of the second chart, which can be the result of bucketing and it may not be that surprising after all.

You can check the accuracy of the graphs directly yourself:

They are screenshots of the linked Google sheet, which has the raw data and graph formulas.

Note there is one "sheet" per diagram (sheet switcher at the bottom inside Google sheets).

Looking at the first sheet, it doesn't seem to be cumulative in the sense of a running sum. If it were cumulative, the first values would be 0, 1, 3.

> If it were cumulative, the first values would be 0, 1, 3.

No.

When you count events, each one contributes a count of 1. The running sum of that is sequence 0, 1, 2, 3 ...

The meaning is "how many job postings have posted until time T.

Reply on news.ycombinator.com