Introduction to Statistics

Part 1

My example of quality graphs comes from the well known Anandtech website. This site does in-depth review of technology, (more often than not over my technical expertise), but the results are presented in an easy to read format. For example the graphs here: http://www.anandtech.com/show/10084/the-toshiba-q300-ssd-review/3 are straightforward to see what objects are being graphed. In this case solid state hard drive data previously compiled had a new drive on the market added to the list. The data is zero based so it doesn’t throw a skewed visual comparison, and gives both an estimation and actual value for each data point. The color coding of the new Toshiba SSD is a nice touch to quickly see where the new drive stacked up amidst the competition.

Part 2

So, I find it's actually difficult to find a bad chart or graph unless one is staring at you. You have to think of places you might first find a graph and often times the worst ones you don't have access too - for example: poorly constructed polls you see on TV programs. Good luck getting a capture off your DVR or finding the poll data graph online. What I did find was an interesting example of someone else finding bad reporting in a graph from what you would think is a trusted source - the NOAA.

Concerning tornados this graph(http://www.ncdc.noaa.gov/sotc/tornadoes/201513) seems to depict a trend of increasing tornado activity. The discrepancy exists as this report now includes EF-0 sized tornados which are much smaller and were not previously reported historically. When looking at major tornado activity (http://www.ncdc.noaa.gov/climate-information/extreme-events/us-tornado-climatology/trends), the trends have largely stayed the same. The NOAA does account for the change in the body of the article, but the graph should reflect the addition of the new size. I would either overlay the EF-0 data in another color, or start a new graph that only goes as far back as good EF-0 data does. Obviously, the changing definition of “tornado” skews what was a statistically normal graph into a right-skewed graph.

The original article pointing out these discrepancies can be found here: https://notalotofpeopleknowthat.wordpress.com/2016/03/11/noaas-misleading-tornado-graph/

2-

The Twitter’s Most Talked About Shows chart represents an example of a good chart.  This chart has specific data presented in a simple and easy understood manner.  The only change I would make would be putting the number of tweets on the y axis and the name of shows on the x axis, so it reads more like a traditional graph.   

Twitter’s Most Talked About [Nominees for Outstanding Drama Series 2012] Shows

At first glance I thought this Food Bank Use in the UK chart, was a great example.  However, an issue I observed was the years listed on the x axis.  It is not possible to determine the dates data was gathered to know if it is being accurately represented.  Is it June of 2008 to June of 2009. Is it December of 2008 to January 2009?,etc.  The exact dates of data collection is important to verify the information was not manipulated.

In the World Education Chart, there is too much going on.  This scatterplot chart does not effectively communicate information to the reader.  Expecting readers to interpret population numbers by use of different size circles, then having to decipher each color, while comparing it to the length of years in school and GDP, all make this graph very difficult to understand and even more difficult to get meaningful information from.  In addition, the use of 1 million, 50 million 100 million, and 1 billion as classes leaves a huge margin between class boundaries. For example, if the population was 25 million, would it be rounded down to 1 million or rounded up to 50 million.  Instead, this should have clearly identified lower and upper class limits.