# Visualization

## Question:

Who benefits more in (opposite sex) marriage?

This is a very broad question which can be approached form many directions. There are entire books written on the subject of marriage and its benefits. To tackle such a question it needs to be greatly refined and narrowed down. As such I narrowed the question to:

What effect does (opposite sex) marriage have on mortality for men and women?

The chart bellow shows the distribution of ages at death for married and single men and women in England and Wales (in 2007). An important point to note is that the values on the graph represent a percent of total deaths which makes the vertical values largely meaningless in absolute terms – this is done to point the viewers’ attention on the distributions as opposed to the absolute number of people in each category.

Caption: While the direction of causality is unclear, it does appear that married men live three to six years longer than single men. Women's longevity, on the other hand, does not seem to be affected by marital status.

## Analysis:

It is immediately obvious that the female graph has very similar distributions for the married and single women; both peak at 84 and almost completely coincide past 90 – this suggests that women are not affected by their marital status and are about as likely to die weather single of married. The male graph, on the other hand, tells a different story; not only are men more likely (by a factor of 2) to die in their youth, a well known fact, but also their married/single distributions are more dissimilar. It can be clearly seen that the most likely time to die for a single man comes 3-6 years before his married counterpart. This graph therefore suggests that marriage has a positive effect on male longevity while having no effect, positive or otherwise, for women. This, in turn, would suggest that men benefit more from marriage than women; at least in terms of lifespan.

# Notebook:

Bellow is the rough outline of the design process.

## Inspiration:

The inspiration for this question came out of an argument of the benefits of marriage.

## Getting data:

The data was obtained from www.statistics.gov.uk more specifically the 2007 repost DR_07_2007.pdf which contained in it the raw data table: Deaths: age, sex and marital status, 2007 (http://www.statistics.gov.uk/downloads/theme_health/DR2007/Table4.xls). British data was used due to the convenience and competency of statistics.gov.uk and due to the fact that marital status is recorded upon a person’s death (in England and Wales) making these statistics available. I suspect that these findings will be mirrored in US data.

## Formatting data:

The data was already very nicely formatted:

To get this data into Tableau readable format some labels were re-worded, all the totals/aggregate/blank rows and columns were removed and all the zero data marked by “-“ was converted to 0s.

## Deciding how to plot:

The marital status at death is entered by the coroner into Form 310

As one of five statuses: Single, Married, Widowed, Divorced, and Not Known.

My first job was to combine these together into two calculated fields. One field to represent the singles and one filed to represent the married population. Now it is a known fact that women, on average, outlive man.

As expected, and from the graph above, most men die married while most women die widowed. Therefore the “Widowed” status has to be in the same category as “Married”.

From the description of the form in the document I understood that the “Not States” status is given in the rare case where it was impossible to identify the deceased, on the assumption that this would really happen to a married person I decided to add it to the “Singe” bin (not that it matters much as the “Not Stated” people account for a tiny amount of the dead).

Lastly, the “Divorced” status was a bit tricky because the term is not well defined. I decided the stick it into the married group making it “Has been married at some point”. Had I done otherwise the graph would have ended up looking like this:

Next I needed to figure out how the graph would be drawn. A simple line graph seemed like a good idea as it represents the rate of change, but I also experiments with other designs like:

The above graph amused me to no end but was, unfortunately, a horrible and unclear way to represent the data. The man/woman figures did give me inspiration for the final design.

Bellow is an example of one more failed attempted to best a line graph:

Finally I decided to bin the age into bins of size 3 to increase the resolution as well as to filter out ages bellow 21 to focus the viewers attention on the interesting part. To highlight the distributions – as opposed to the magnitude of the values – “percent of total” scaling was used on the data. Finally I observed that having two graphs looks much clearer than having one combined graph, and decided to stack them on top of each other and use the same age scale to aid comparisons. The female graph went on top to be the setup and the male graph took its place as the punch line.