Databases Reference
In-Depth Information
values have already been handled. For the most part, the columns with missing
values were eliminated.)
View UtHomesAsDownloaded.csv in a scatter plot.
Select Longitude for the X axis and Latitude for the Y axis.
Rotate as needed to view the height densities.
A scatter plot with longitude and latitude on the axes is essentially a location
plot without the background map layer. A scatter plot was selected because of
its ability to plot density histograms (Type
Height Density). In reviewing
the plot, notice the highest concentration of homes for sale around latitude
40.406 and longitude
111.883. Given domain knowledge of the metropolitan
area, the validity of homes concentrated in that small area was questioned.
Using the parallel plot and the text-based table view, we zoomed in on those
homes to find 123 entries with the exact same latitude and longitude; yet the
street addresses were all different. Apparently, the realtor entering the listing
did not have the latitude and longitude of the property and did not bother
looking it up. Instead they entered a generic location. In checking other areas
of high concentration, the same problem was found although not to the
same extent.
To correct the problem outside of VisMiner, publicly available geocoding
services were used. Geocoding is the process of converting a street-city-state-
zip address into latitude and longitude coordinates. Map serving sites such as
Google, Yahoo, and Microsoft include free geocoding services in limited
quantities. The University of Southern California GIS Research Laboratory
(
webgi s.usc.edu
), whic h was used for thes e entrie s, provides fre e geo coding for
up to 2,500 addresses. It applies usage-based fees beyond that level.
ΒΌ
Distribution consistency
The dataset BodyTemp.csv contains readings of patient body temperatures as
collected by nurses in a hospital over a two-week period.
In the dataset BodyTemp.csv, view the temperature column distribution in
a histogram.
The temperatures are recorded as integers in tenths of a degree Fahrenheit.
In the distribution, we see the opposite of what we saw in the UtHomes dataset.
Instead of an unusually high concentration of observations, we see a lower than
expected number of normal (98.6 F) patient temperatures. Upon investigation,
the hospital found that nurses were reluctant to report a temperature of 98.6 as
it might lead a supervisor to question whether the temperature had actually
