4 ms·

These charts show how statistics can be used to mislead. The scales on these are all out of whack. If I am reading the scales properly, the salary difference b

These charts show how statistics can be used to mislead. The scales on these are all out of whack. If I am reading the scales properly, the salary difference between product managers and software engineers is about 10k a year in salary, out of 120k-ish. That's roughly eight percent of an already high salary. What sort of sampling are they using, just the companies that use an online job board? Do the companies using "Hired" to give offers consist of a representative sample, or do they skew towards small companies, or companies where software is considered a cost center instead of a revenue center?

forgetsusername 10y ago
>These charts show how statistics can be used to mislead. Someone says this every time a chart with a scale other than zero shows up somewhere. It's untrue, generally. The context of the information is what is important. In this case, what is being emphasized is the difference between salaries of positions. What would the point be of having a mass of whitespace at the bottom of the chart? What relevance is the salary level in relation to zero? Scaling to zero would diminish the information being conveyed.
- jimmywanger 10y ago
  No, you're saying "emphasized". It's more exaggerated. When our lizard brains look at charts, we're conditioned to check how relatively apart the lines are. By removing the zero scale, we're diminishing the lack of the absolute proportions of the salary scales. 8% doesn't seem like much, but 10k seems like a whole lot. Also, this is a sensationalist article. They did not get into their methodologies, or whether their data points were actually representative of the industry. Somebody getting paid by the column inch found an easy article to write.
  - forgetsusername 10y ago
    Edward Tufte himself says to use a non-zero axis for time series. If the extent of your analysis is "10k is a lot, 8% isn't" then I'm not really concerned how you interpret my chart. I have no comment on the quality of the article.

forgetsusername 10y ago

>These charts show how statistics can be used to mislead. Someone says this every time a chart with a scale other than zero shows up somewhere. It's untrue, generally. The context of the information is what is important. In this case, what is being emphasized is the difference between salaries of positions. What would the point be of having a mass of whitespace at the bottom of the chart? What relevance is the salary level in relation to zero? Scaling to zero would diminish the information being conveyed.

jimmywanger 10y ago
No, you're saying "emphasized". It's more exaggerated. When our lizard brains look at charts, we're conditioned to check how relatively apart the lines are. By removing the zero scale, we're diminishing the lack of the absolute proportions of the salary scales. 8% doesn't seem like much, but 10k seems like a whole lot. Also, this is a sensationalist article. They did not get into their methodologies, or whether their data points were actually representative of the industry. Somebody getting paid by the column inch found an easy article to write.
- forgetsusername 10y ago
  Edward Tufte himself says to use a non-zero axis for time series. If the extent of your analysis is "10k is a lot, 8% isn't" then I'm not really concerned how you interpret my chart. I have no comment on the quality of the article.