OmCheeto
Gold Member
- 2,482
- 3,408
0.026Ygggdrasil said:Just looking at the graph, it looks to me like there is a negative correlation overall (have you calculated the correlation coefficient for the data?).
(my apologies for taking so long. I'm fairly certain I've never run across "correlation coefficient" in my studies, and had to figure out what it was. I used the first wiki reference I could find.)
Not sure if I did this right.Vanadium 50 said:I think the better thing to do is to sum this into deciles
I'm 99.9% confident that the maths is correct. My spreadsheet has a "Show R² Value" function when I make it display a linear trendline and the "r²" values match my "r" values in all 10 cases.
As to how to interpret this bar-graph, um, I'm going to need some assistance here.
It looked to me like decile #3 has a really nice "r" number(0.76). So I removed just one of the data points(Bronx NY) and "r" went down to 0.30. Removing the next two(Philadelphia PA and Nassau NY) changed "r" to -0.15. I stuck me as odd that 3 entries out of 25 could have such a large effect.
In other words, do you analyze each individual decile, or do you look at them as a collective?
Or perhaps groups? Deciles 4 and 5 seem to be not that different in population sizes, yet the coefficients are polar opposites. Ditto with deciles 3 & 4, 6 & 7, and 1 & 2.
