Charts

• We will be working with this data.
  • Housing sales in 2014/5 in Kings County Washington State.
  • From Kaggle.
• Setup
  • Delete the ID column.
  • Please add three columns after A and turn the date field into Year, Month, Day
    • Do this with text to columns.
    • But this time, use a fixed width field.
    • Put a line after the first 4 digits (the year)
    • And after the next two.
    • And after the next two.
    • 
    • Add labels
    • Delete the new column d.
    • 
  • Select the entire table and use the name manager to name the columns.
• If you understand the basics of charts they can povide insight into the data more than any thing else.
  • I tend to be visual
  • So I create a buch of charts.
• In data science there are frequently two stages of making charts
  • Exploratory analysis
  • Presentation.
• You find MANY documents that discuss the difference.
  • Don't feed people a bunch of charts from data analysis.
  • These are your work, to figure out what is happening in the data.
  • These don't need to be completely documented.
  • But they do need to be accurate.
• For presentations
  • I love the book storytelling with data by Cole Nussbaumer Knaflic.
  • But This article will do.
• The rules for presentation graphs
  • Everything is labeled
    • Title
    • Axis
    • Data
    • Correctly labeling your data helps
    • Add titles and axis labels and everything else.
    • This is a matter of discipline.
    • It is way too easy to just slap up a chart and consider it done.
  • Make sure it is right
    • And don't change scales or other items to make it "better"
  • Don't go wild with chart types
    • Use charts people understand.
    • Avoid 3D
    • Make simple charts/graphs.
    • Eliminate clutter.
  • Some of what we will do in the next will contradict this.
    • For exploration we will probably not label everything.
    • We will probably overdo charts as well.
    • We want to learn as many pieces as we can.
    • But just because you know them doesn't mean you need to use them.
• Just a comment
  • The charting tool is very extensive.
  • There are hundreds of options for a chart.
  • And even if I knew them all (which I don't)
    • We would not have time to cover everything.
    • You would not remember everything.
  • In this area you need to
    • Have some vision for what you want to produce.
    • Be aware of the options as you attempt to "modify" the chart.
    • Be persistant, use resources, and make the details of the chart look the way you want it to look.
  • Remeber, you are the master of your work, not chained to the default settings in excel.
• Let's try a graph.
  • My intuition tells me that the size of the house is related to it's price.
  • So let's build a scatter plot to see if that is true.
  • Insert a new worksheet
  • Select the Price column (D) and the sqft_living(G)
  • Note, I have recycled these notes, so some of the columns in the pictures may be off
  • On the insert tab, select the scatter plot chart
  • 
  • I cut and pasted this to the new worksheet.
  • 
• What is a scatter plot?
  • For each pair (price, sq_ft) it plots a point.
  • It does this for all of the data.
  • So what do we see in this scatter plot?
    • There is a correlation between these two values.
    • Correlation does not imply causation.
    • And it is perhaps not quite as strong as we would like
• Let's try an experiment.
  • Grab a new worksheet
  • In column A generate the numbers 1 = 200 (just to keep track) in a column
  • In columns B and C generate 200 random numbers with =randbetween(1,1000) in the next two columns.
  • Copy these to columns E-G, paste as values
  • Generate a scatter plot from columns F and G
  • Let's try something silly.
    • norm.inv will give us an inverse normal function.
    • We can use this to generate normally distributed data.
    • In J2 put x, K2 100
    • In J3 put s, K3 10
    • In H2 put =F2+norm.inv(rand(),$K$2,$K$3)
    • 
    • plot this (columns F and H)
    • Change s to be 50
    • Change s to be 1000
  • Sometimes it is helpful in a scatter plot to see a trendline
    • Click on the chart, the + sign and select trendline
    • 
    • Click on the arrow and select more options
    • If you are not on the Trendline Options, select these.
    • 
    • Select liner
    • Also select show trendline.
    • Move to the line section
    • 
    • Select red (so we can see the line)
    •