Talk:Gini coefficient

Formula for the Gini coefficient

It seems to me the current formula in the article might be slightly incorrect. Here is the current formula:

$G={\frac {\displaystyle {\sum _{i=1}^{n}\sum _{j=1}^{n}\left|x_{i}-x_{j}\right|}}{\displaystyle {2\sum _{i=1}^{n}\sum _{j=1}^{n}x_{j}}}}={\frac {\displaystyle {\sum _{i=1}^{n}\sum _{j=1}^{n}\left|x_{i}-x_{j}\right|}}{\displaystyle {2n\sum _{j=1}^{n}x_{j}}}}={\frac {\displaystyle {\sum _{i=1}^{n}\sum _{j=1}^{n}\left|x_{i}-x_{j}\right|}}{\displaystyle {2n^{2}{\bar {x}}}}}$

Specifically, the current numerator suggests it would include each person's difference from themself, i.e., $x_{i}-x_{j}$, where $i=j$. However, if $j$ is defined as the set of all other persons who aren't $i$, then the upper limit of $j$'s index ought to be $n-1$. Thus, I believe the numerator ought to be:

$\displaystyle {\sum _{i=1}^{n}\sum _{j=1}^{n-1}\left|x_{i}-x_{j}\right|}$

Moreover, I propose that the article be updated to explicitly define $j$.

Lastly, I'm not sure whether the denominator needs any updating. MosesRivera100 (talk) 19:07, 7 September 2023 (UTC)

These formulas added later may be wrong. Take $x_{1}=1;\;\;\;x_{2}=0$

Using these would give: $G={\frac {1}{2n^{2}}}{\frac {\sum _{i}\sum _{j}|x_{i}-x_{j}|}{\bar {x}}}={\frac {1}{8}}\times {\frac {|1-0|+|0-1|}{0.5}}={\frac {1}{2}}$

Now use $G={\frac {\text{mean absolute difference}}{\text{mean}}}={\frac {\bar {\Delta }}{\bar {x}}}$

Where ${\bar {\Delta }}={\frac {1}{n}}\sum _{i}\sum _{j>i}|x_{i}-x_{j}|={\frac {1}{2n}}\sum _{i}\sum _{j}|x_{i}-x_{j}|$

and ${\bar {x}}={\frac {1}{n}}\sum _{i}x_{i}$

You'll get the correct value of 1. In this case, $G={\frac {\sum _{i}\sum _{j}|x_{i}-x_{j}|}{2\sum _{i}x_{i}}}$ 2804:1998:421:F201:88E7:6514:10BF:CCDF (talk) 19:57, 25 June 2024 (UTC)

Using same projection for wealth and income Gini maps

Using the the same projection would allow for easy comparisons between wealth and income Ginis Mrsmrmrmrs (talk) 13:55, 18 January 2024 (UTC)

• What I think should be changed: Although both Gini coefficients are defined as areas between certain curves and share certain properties, there is no simple direct relationship between the Gini coefficient of statistical dispersion and the Gini coefficient of a classifier.
• Why it should be changed: Both Gini coefficients are defined as areas between certain curves, and in the case of known ratio of "event" to "non-event" in the data there is a simple relationship between them. G_t = G_n/(1+α). G_t is the traditional measure of statistical dispersion, G_n is the new measure for binary classifier, and α is the ratio between events to non events in the data.
• References supporting the possible change (format using the "cite" button):On the connection of two GINI coefficients by Michael Roginsky https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4715030

Rogmike (talk) 19:11, 27 February 2024 (UTC)

References

 Not done: Rogmike, source displays the following: SSRN Search Results

This paper has been removed from SSRN at the request of the author, SSRN, or the rights holder. ARandomName123 (talk)^{Ping me!} 01:45, 4 March 2024 (UTC)

ARAndomName123,

Thank you for looking into my minor edit. the article I put on SSRN was taken out due to formality (they need my name on the first page of the PADF file). I will resubmit it shortly. If you prefer I can send it to you directly, so you can check the formulas (rather tedious but elementary).

Rogmike Rogmike (talk) 16:46, 4 March 2024 (UTC)

ARandomName123, The article is back on SSRN website, same abstract #. Rogmike (talk) 16:37, 5 March 2024 (UTC)

The calculation section is more confusing than...

Mathworld's article on the Gini Coefficient: https://mathworld.wolfram.com/GiniCoefficient.html Am I missing something? This wiki article makes it seem like the Gini coefficient is difficult to use and fraught with pitfalls without giving many concrete definitions, while the Mathworld reference has two easy-to-understand formulas. What's up with that? 24.228.174.225 (talk) 21:51, 5 August 2024 (UTC)