18 ms·

That's why I cited median values - so your complaint does not apply, since by mathematical fact a tiny minority cannot change the median appreciably.

coliveira 4y ago
People don't seem to understand that median is not a measure that is much better than mean, especially in the case of high inequality. For example, if I take all the money from the 40% poorer population and distribute it to the 40% richer population, the median is still the same. If you don't believe, just compute the mean and median of these two sequences: [1,2,3,4,5,6,7,8,9,10] [0,0,0,0,5,6,11,11,11,11] Another example that I know fully well is Brazil: the mean is 8,560 BRL and the median is 8,220 BRL for monthly salaries, that is, very close. But these monthly salaries are known to vary widely from 2,170 BRL (USD 417) to 38,200 BRL (USD 7349), so hardly anyone is really making 8220 or close to that.
- ChrisLomont 4y ago
  >People don't seem to understand that median is not a measure that is much better than mean, especially in the case of high inequality. Yes, they do. I am a PhD mathematician with decades of work using these tools. I am quite clear on the meanings and shortcoming of all sorts of indicators. I am also aware how real world data trends instead of cooked up examples that don't match reality. Median is a common value used by economists precisely because it more likely shows what a person makes as opposed to the mean, especially in cases of high inequality. > If you don't believe Yes, you can make up all sorts of nonsense. Not a single country follows the distributions you made up. >Another example that I know fully well is Brazil: You conclusion does not at all follow from your claim. Or from the actual data. > so hardly anyone is really making [the median] or close to that Well, by definition, over half of people make the median or more. Is that hardly anyone? If you look at the data [1] (income by decile in Brazil), besides the 50% making median income or higher, another 30% of the people make within 30% of median. That leaves 20% of the people making under 70% of the median income. What is your definition of "hardly anyone"? Look at the data. Clearly median is much more accurate than mean for explaining how much people make, as this example showed. [1] https://www.statista.com/statistics/1251075/average-monthly-income-percentile-brazil/ https://www.statista.com/statistics/1251075/average-monthly-...
  - coliveira 4y ago
    The two examples show exactly the problem of the median: it is not that different from the mean in many situation when we talk about salaries and incomes. Now, if you start using the percentiles as you're doing here, then you are looking deeper into the data and your results get more accurate and reflective of reality, which just proves my point.
    - ChrisLomont 4y ago
      >The two examples show exactly the problem of the median: it is not that different from the mean in many situation when we talk about salaries and incomes ??? Did you compute them? do you know how? The Brazil data I just posted has median around 1383, mean of 2308.36, which is 65% larger. An interesting test: compute with details the median and mean of the data I just posted. I already gave you the answers. I think you don't understand these terms. That you don't like the outcome of the most two commonly used indicators is not the fault of the indicators - it's the result of you wanting to put things in the data that are not there. Your claim about Brazil having "hardly anyone", which I showed was nonsense, shows exactly what you're trying to do - put things in data that simply are not there (and in the case of your claim, are mathematically invalid no matter the income distribution). >Now, if you start using the percentiles as you're doing here, then you are looking deeper into the data Percentiles have the exact same graph as dollars, or pesos, or anything you desire. % = exact$ / total $. So they have the exact same graph, merely a different label. Do you understand this? The % of total income versus income shows nothing different, there is no "deeper into the data" or "more accurate and reflective of reality" for this. It's simply scaling ALL the numbers by a constant. No one changes place. No income compared to another changes in ratio. There is no subtle nuance pulled out of anywhere. >which just proves my point. Is your point that you don't know what any of these terms mean? Then yes, you proved your point. Mean, median, percent, ratio, decile, all the terms related to income - you have screwed them all up beyond reason. Don't bother replying until you answer the questions I poste last time: Do you admit 50% of people in Brazil make median income or more? What amount of people did you mean by "hardly anyone"? Also answer: is the graph of income by % the same geometric shape as income by $, or yen, or peso, or any currency you care to pick, thereby giving no "looking deeper into the data"?

coliveira 4y ago

People don't seem to understand that median is not a measure that is much better than mean, especially in the case of high inequality. For example, if I take all the money from the 40% poorer population and distribute it to the 40% richer population, the median is still the same. If you don't believe, just compute the mean and median of these two sequences: [1,2,3,4,5,6,7,8,9,10] [0,0,0,0,5,6,11,11,11,11] Another example that I know fully well is Brazil: the mean is 8,560 BRL and the median is 8,220 BRL for monthly salaries, that is, very close. But these monthly salaries are known to vary widely from 2,170 BRL (USD 417) to 38,200 BRL (USD 7349), so hardly anyone is really making 8220 or close to that.

ChrisLomont 4y ago
>People don't seem to understand that median is not a measure that is much better than mean, especially in the case of high inequality. Yes, they do. I am a PhD mathematician with decades of work using these tools. I am quite clear on the meanings and shortcoming of all sorts of indicators. I am also aware how real world data trends instead of cooked up examples that don't match reality. Median is a common value used by economists precisely because it more likely shows what a person makes as opposed to the mean, especially in cases of high inequality. > If you don't believe Yes, you can make up all sorts of nonsense. Not a single country follows the distributions you made up. >Another example that I know fully well is Brazil: You conclusion does not at all follow from your claim. Or from the actual data. > so hardly anyone is really making [the median] or close to that Well, by definition, over half of people make the median or more. Is that hardly anyone? If you look at the data [1] (income by decile in Brazil), besides the 50% making median income or higher, another 30% of the people make within 30% of median. That leaves 20% of the people making under 70% of the median income. What is your definition of "hardly anyone"? Look at the data. Clearly median is much more accurate than mean for explaining how much people make, as this example showed. [1] https://www.statista.com/statistics/1251075/average-monthly-income-percentile-brazil/ https://www.statista.com/statistics/1251075/average-monthly-...
- coliveira 4y ago
  The two examples show exactly the problem of the median: it is not that different from the mean in many situation when we talk about salaries and incomes. Now, if you start using the percentiles as you're doing here, then you are looking deeper into the data and your results get more accurate and reflective of reality, which just proves my point.
  - ChrisLomont 4y ago
    >The two examples show exactly the problem of the median: it is not that different from the mean in many situation when we talk about salaries and incomes ??? Did you compute them? do you know how? The Brazil data I just posted has median around 1383, mean of 2308.36, which is 65% larger. An interesting test: compute with details the median and mean of the data I just posted. I already gave you the answers. I think you don't understand these terms. That you don't like the outcome of the most two commonly used indicators is not the fault of the indicators - it's the result of you wanting to put things in the data that are not there. Your claim about Brazil having "hardly anyone", which I showed was nonsense, shows exactly what you're trying to do - put things in data that simply are not there (and in the case of your claim, are mathematically invalid no matter the income distribution). >Now, if you start using the percentiles as you're doing here, then you are looking deeper into the data Percentiles have the exact same graph as dollars, or pesos, or anything you desire. % = exact$ / total $. So they have the exact same graph, merely a different label. Do you understand this? The % of total income versus income shows nothing different, there is no "deeper into the data" or "more accurate and reflective of reality" for this. It's simply scaling ALL the numbers by a constant. No one changes place. No income compared to another changes in ratio. There is no subtle nuance pulled out of anywhere. >which just proves my point. Is your point that you don't know what any of these terms mean? Then yes, you proved your point. Mean, median, percent, ratio, decile, all the terms related to income - you have screwed them all up beyond reason. Don't bother replying until you answer the questions I poste last time: Do you admit 50% of people in Brazil make median income or more? What amount of people did you mean by "hardly anyone"? Also answer: is the graph of income by % the same geometric shape as income by $, or yen, or peso, or any currency you care to pick, thereby giving no "looking deeper into the data"?