View source

=Datasets=
For this assignment, I used data sets on education and sexual health issues. I found these datasets via the excellent [http://data.un.org UN Data] search utility.

=Initial Question=
* How are education indicators related to sexual health issues (e.g. contraceptive use and the legality of abortion)?
** ''As might be expected, I'm coming in expecting better education statistics (such as literacy rates) to be positively correlated with greater contraception use and less strict abortion laws.''

=Notebook=
==Iteration 1==
I began by pulling in two datasets, with literacy rates and contraceptive use data, and joined the imported tables in Tableau. Looking for a relationship between two variables, the obvious starting point was a scatterplot:

[[Image:contraception_scatter.jpg]]

The scatterplot seems to show the expected relationship between literacy rate and contraception use. For kicks, I also imported a dataset with primary school enrollment rates and produced the same scatterplot with these data instead of the literacy rate data. Again, there is a trend of increasing contraceptive prevalence with better education measure.

[[Image:enrol_contraception_scatter.jpg]]

==Iteration 2==
Next I pulled in data on the legality of abortion on different grounds in various countries. I wondered whether there was a relationship between the prevalence of contraceptive use and the number of legal grounds for abortion.

[[Image:abortion_contraception_bar.jpg|800px]]

This visualization shows no clear relationship. There is some indication that the more lax the abortion laws, the greater the rate of contraceptive use is likely to be, given the more consistent thickness of the contraceptive prevalence bars towards the bottom. But the appearances do not suggest much. Row reordering within each bin of legal reason counts (in [1, 7]) might make any relationship clearer, but in the next iteration I think I used a better approach, since the per-country information was not what I was interested in and its appearance in this visualization adds to the clutter.

==Iteration 3==
Wanting to pursue the question from iteration 2 a bit more, I put together the following visualization, which gives an indication of the conditional distribution of contraceptive prevalence given each 'level' of abortion legality, that is, for each number of legal grounds for abortion in a country.

[[Image:contraception_distribution_given_abortion.jpg]]

Presented in this way, in a small multiples layout, we can discern clear differences in the conditional distributions of contraceptive prevalence, and we see that for countries where there are 7 legal grounds for abortion, the distribution of contraceptive prevalence is shifted to the right. At all other levels, the conditional distribution is fairly uniform but shifted toward the left.

=Discussion=
My initial question was a bit broad, and I ended up taking the investigation in a different direction. The final question I ended up answering, whether there is a relationship between contraception use and abortion laws, was more interesting and the visualization suggesting the answer exemplifies some of the key lessons we've learned so far. The 'answer' still leaves much unexplained, but being able to even get this far so easily is impressive. Additionally, playing around with Tableau and creating even useless (for this assignment, at least) visualizations was great fun.