Workbook is attached, thank you.
1 of 1 people found this helpful
Quite a bit challenging.
Even only for Trend line, needed to duplicate Data source.
I used Union.
Else, several trick are needed. I don't have time now and explain later if you want.
1 of 1 people found this helpful
May I offer something not related to your question?
Please find the attached with Sheet 7 as an example.
The main idea is to use distribution bands (Quartiles) instead of Trendlines.
It's a feeling that even if Gini and GDP per capita are related (within WB Groups or not),
their relationship is far from being modelled by a linear (or other kind of) regression.
The distribution of the dots (Country-Years) on a scatterplot gives more food, imho.
As one could see, both Low and High WB Groups have more "compact" Gini scores
(the interval between the Lower & the Upper Quartiles, where 50% of the dots are) --
as compare to the whole population (or to the LM & UM groups having wider inter-quartile ranges).
This could be explained by a common-sense-grounded hypothesis
that with growing GDP per capita there are opportunities for men in power
to re-distribute the higher amount / proportion of common wealth --
thus leading to higher inequality. On the other side, the H-Countries
may have been developed institutions & instruments to re-distribute
common wealth "back" (to their "lower-income" social groups).
Another obvious viz option would be placing [Year] on Pages shelf
and letting Gapminder-style analyses flow. With that one could easily see
some Countries to come to a higher GDP per capita starting from 1999-2000,
and some of them about to change their WB Group on the way.
Though their Gini scores did not change so much.
PS Besides, I'm not fond of Gini and GDP per capita
as the metrics for inequality and wealth, correspondently.
I'm well aware that we would be waiting for better ones forever.
Very interesting stuff, thank you.
This was exactly what I needed, and one more additional category would have made it perfect:
You currently have 6 categories on your lookup filter: All, Dot Only, H, L, LM, and UM. If there is one more regression line for all the dots (think of it as the regression line for "world") that would complete the task.
But regardless, this was very insightful and helpful. Let me look into the worksheet and see if I can replicate this, it will be very useful for my future work.
Thank you Yul for your suggestion, this was also an interesting visualization. The use of quartiles is a good idea, I might just use it in a future viz.
A thing to note is that there might be a bias with the income groups (such that high and low groups have less countries, hence less variation could be seen) but I am not sure, so have to explore that. This is supported by the idea that I used time-varying income groups, which essentially implies that the end-groups may intuitively have less countries than the ones in the middle categories.
But as you note, Gini and GDP may not exactly the best indicators, maybe in the future there will be better metrics that can tell the story.
If there is one more regression line for all the dots (think of it as the regression line for "world") that would complete the task.
In my head, there is no was way to do that. Cause (as long as I know) it requires one more data set which can not be overlayed as third axis.