...to Geog 353 Main Page and Course Description
...to Geog 353 Syllabus
...to Geog 353 Course Schedule and Lecture Outlines
...to Geog 353 Laboratory Information and Student Projects

100 points

ASSIGNED: Monday October 14

DUE: Wednesday October 23: Choropleth ("graduated color") classification

DUE: Wednesday October 30

You have cleaned up your historical Census data and brought it into ArcGIS, where you have linked it to a base map of your state (or states). You have used ArcGIS's spreadsheet / database capabilities to calculate the data on population change. Very exciting.

You will now map the population totals and percent change data using four mapping techniques: the choropleth, graduated symbol, proportional symbol, and dot map.

Lab 7 helps you to choose an appropriate classification scheme for your data - in essence, simplifying it by dividing up the data into a series of categories. You will need to classify data for the choropleth and graduated symbol maps you create. As we are creating a series of twelve maps that will be animated, you must develop a single classification scheme for the total population and percent change data.

Lab 7 Goal: Selection of a single classification scheme (number of classes and data divisions) for your choropleth and graduated symbol maps; generate these maps as well as proportional and dot maps for your Census data.

The Details:

To complete this exercise, you must be familiar with the way that ArcGIS classifies data, so take some time to familiarize yourself with these options and how they work. A basic overview of ArcGIS's classification capabilities was discussed in the Data Classification lecture, and the on-line help (in ArcGIS) or your kindly instructor can probably answer any other questions you may have.

We will use four different mapping types (to be reviewed in the upcoming Map Symbolization lectures) to map our historical population data: the choropleth map (called graduated color in ArcGIS), the graduated and proportional symbol map, and the dot density map. The choropleth and graduated symbol maps require your data to be classified. The proportional symbol and dot maps do not. The initial part of this exercise focuses on classifying your data.

You should have your ArcGIS file (.mxd) set up with your state or states linked to the table which includes your historical Census population totals and percent change prior to beginning this exercise.

Make sure you have exported your data, so you have only your state (or states) map and a data table that is no longer virtually joined. We did this in the last exercise (step 8 in Lab 5).

We will begin by creating a series of choropleth maps, which in ArcGIS are called Graduated Color maps.

1) You should have one layer in your data frame: mine is called Wisconsin2. Change the name of this layer and generate a choropleth map of one decade of population change data.

• Double-click on the layer: this is a short cut to the layer properties (also accessible by a right click on the layer, then select layer properties)

• Under the General tab, type a more appropriate Layer Name: such as Population Change 1900-1910

• Select the Symbology tab: this is where you adjust the symbolization of your maps. In the list on the left you should see the term Quantities and under that, the four mapping types we will use: graduated colors (choropleth), graduated and proportional symbols, and dot density (dot map). Click once on Graduated colors

• You must tell ArcGIS which field (column) of data you want to map. The first map to create is the population change from 1900-1910. In the Fields box, next to Value: select the appropriate field from your table: the 1900-1910 population change data. Leave Normalization as none A default classification scheme and color scheme will appear. We will adjust this later: for now hit OK

• You should see a choropleth map of the population change in your state or states for 1900-1910. In the table of contents, under your layer name, is the name of the field. Click once, then again on the name to edit it, then type a more appropriate name for this set of data: again, Population Change 1900-1910 is a good choice.

2) Now that you have one choropleth map on a layer, copy it and modify it for the next map of population change data, 1910-1920:

• Right-click once on the layer Population Change 1900-1910

• From the menu select Copy

• Right-click once on the data frame (mine is called Wisconsin Dataframe) and from the menu select Paste Layer. The pasted layer shows up at the top of the list: you can arrange the order of the layers by clicking and dragging. Leave the layer where it is for now.

• Repeat the process you followed in step 1 above to modify the name of the layer and the data (field) so that the 1910-1920 population change data is mapped.

• Do this for the rest of your population change data: you should have twelve appropriately named layers, each with the correct data selected and mapped.

3) Get rid of data values that are the result of division by the 9999999999 values in your data set. Recall we used those nines to indicate years where there was no data. You don't want to use those numbers in your classification as they are not real data. The numbers you want to exclude in the percent change columns are either -99.9999 or -100 or possibly a zero. A few of you had odd things happen to your 9999999999 values - such changing to 100000000000. You thus may have different numbers: but in any case, you have to exclude these "missing data" values.

• Right-mouse-click on your first layer, go to Properties then Symbology window. Click on the Classify button then the Exclusion button.

• Exclude counties where the value is equal to either -99.9999... or -100 or maybe even (gasp) zero - it depends on the specific number that was calculated (it may vary from student to student). You can exclude any number lower than -90, for example, if there are no real values that low in your data.

• Repeat this process for any other layers that have fake numbers. You may have to go back and check that the appropriate values are eluded later - after you do step 7 below. Also, when you are mapping out totals, you will have to exclude the 999999999 values.

• Think! Make sure you are not excluding any real numbers. For example, if you elude all numbers less than (<) -99, and you have a real value in your data of -137 percent change in population, that number will get exluded. Badness! Figure out how to exclude only the values that need to be excluded.

Now you have to classify the data.

4) Choosing a specific classification for your twelve maps is going to require thinking and work. Refer to the Data Classification lecture and the Making Maps chapter on Generalization & Classification: you must choose the number of classes (four, five, or six is normal, but there can be good reasons to do more or less) and the particular classification scheme.. It is unlikely that one of the default classification schemes in ArcGIS (natural breaks, equal interval, quantiles, standard deviation) will work: you will probably have to modify one of them and thus devise one on your own (eg., a custom classification scheme). The reason for this is simple: you are creating twelve maps, and these twelve maps have very different data ranges, and you have to come up with one classification scheme that works for all of them (the lowest and highest values in all your twelve decades of data have to be included in the scheme) so you can compare the maps to each other (necessary as we are going to animate the series of maps).

5) ArcGIS has a very nice way of visualizing data classification. Return to the Symbology window and click on the Classify... button (top left). You can modify the Classification Method and Classes here, and see the resulting breaks on a histogram. Break Values are the specific locations of the breaks in your classification scheme. You can also interactively modify the classification scheme using the histogram: simply grab one of the blue class breaks and move it. This results in a manual classification scheme. To reset the breaks, simply select the appropriate classification scheme in the method window.

6) It is important to devise one classification scheme that fits every one of your twelve maps. You may know this, as I have mentioned it several times. This means that the lowest data value in all twelve fields is the bottom value in the classification scheme, and the highest data value in all twelve fields is the top value in the classification scheme. Thus you have to look at your data. You can do this by viewing your data in a Table. Right-click on the field name to rearrange your data from high to low (sort ascending). Find the highest and lowest values in all twelve fields of data.

Now that you have the low and high end of the classification scheme, start to think about dividing up the data. Natural Breaks is a great default, or you may consider equal intervals or quantiles. But ponder the nature of your data and the phenomena (population change): is there is a significant value in the data set that would make a good initial breaking point? For many of you, there will be both negative and positive values in your data: some counties lost population, others gained population. Thus a significant value is 0. This is a good initial division of your data: losses and gains. This is significant to what you are trying to show on your maps, eg., population change.

What about the other class breaks? This will depend on your data. Look at the range of negative values. If there are few values, and they are relatively small, you may want to have one or maybe two classes of negative values. If there are many negative values, you may want more classes of negative values. The same for your positive values. In the end, you may end up with a classification scheme with one class of negative values, and three classes of positive values:

-12 to 0

0 - 15

16 - 30

31 - 60

Those of you with a larger range of negative numbers may have several classes below 0 and only a few above. Or you may have three below and three above.

Those of you with no negative values have to decide on a classification scheme which best represents the distribution of your positive values. Try several different options.

Write out a few options (you can do these in your lab log entry).

To evaluate the effectiveness of your classification choices, use Open Office to create a histogram of the percentages from all twelve of your data sets. [I recall that it was easier to create a histogram in Open Office than Excel, but you can try Excel if you want].

• Open Open Office and then navigate and Open the .dbf file associated with your percent change data. If you don't know which file this is, right-mouse- click on the layer in ArcGIS with your percent change data and select Properties. Click on the source tab to see the .shp file associated with the layer. The .dbf file has the same name as the .shp file, but ends with .dbf.

• Be careful not to change this DBF file in any way, or save it.

• Copy the percent changes (just the percentages, nothing else)

  In Open Office make a new spreadsheet document

• Paste the percentages into this new document

• Figure out how to create a histogram in Open Office, and look for "natural" breaks or gaps in the data - these can be the breaks in your classification. Ask your instructor to look at your histogram and get some feedback on the gaps you see in the data.

  ---

  Making a histogram in OpenOffice (Julian Kusin & Ebosehon Imeokparia)

  • Create two columns on the left side including all of your data. The first column should have your classes from lowest to highest, the second will be a frequency (the number of data values that will be in each class): creating frequency is detailed below.
    • Add the values for the classes (this is either the year or the time range) in the first column
    • Click to select the top row of the second column
    • Go to the insert menu at the top and click Function
    • Scroll down to Frequency and hit next
    • To the right of "data," click the small button with a green arrow
    • Drag a box from the top left cell of your data down to the bottom right cell
    • Click the green arrow button again
    • Do the same for classes, but start with the top class, down to bottom class
    • Hit OK, and your second column should be filled with the frequency for your classes
    
    
  • To make the actual histogram graph
    • Click on the bar graph button at the top of the window
    • Leave chart type as column, and click "2. Data Range" in the left column
    • Click the green arrow button to the right, and highlight a box from the top left class to the bottom right frequency
    • Make sure the "data series in columns" is selected and the "first row as label" and "first column" as label boxes are checked
    • Hit finish.

  ---

Try your classification in ArcGIS: you can type custom breaks in the Range column of the Symbology window then look at the results on the map, or you may use the interactive histogram. Whatever cooks your egg.

When you come up with a classification scheme, see if similar counties are in the same class. If a classification scheme puts too many dissimilar counties together, it may not be appropriate. Bottom line: you have to look at all your data, from high to low, and think about how best to divide it up. You may end up with four classes, or eight. You may choose equal intervals, or quantiles. But you must be able to justify why you chose the classification scheme you chose. How does it take into account the distribution of the data? How does it help to communicate to people who will see the map the characteristics of population change in your state? Use your brains! This is not a situation where ArcGIS can really help you decide on the best way to classify your data. And, indeed, you should never use the default classification without looking at your data and thinking about the phenomena the data represents. Work with your data, and different possible classification schemes, and document what you did and why you chose one particular scheme in your Lab Blog.

7) Set up your chosen classification scheme for one layer, then you (and only you) can easily impose the same classification scheme on all additional layers by importing the scheme. Please follow the order of instructions below very carefully; if you don't the process won't work and, even worse, you may blow out the Pfeiffer Valve in your computer.

• Select the layer (% chg) with the lowest low value of all twelve of your percent change data columns (fields). You can figure this out with your histogram (from the previous section) or by viewing your table of data.

• Really. Make sure it is the layer with the lowest low value.

• Go to properties for the layer, then the symbology tab. Type in the number of classes you are going to use. Type your class breaks in the Range column, from the low values up, hit return after each correct break value is entered. The high class break or breaks may not work - don't worry for now.

• Save your ArcMap project and hit OK to get out of the symbology window.

• Select the layer (% chg) with the highest high value of all twelve of your percent change columns (fields).

• Go to properties for the layer, then the symbology tab. Hit import. In the window that appears, select the layer you just set your classification scheme up in (the one with the lowest low value). Hit OK.

• In the next window, make sure the data field is the correct data field for the layer you are working on (not the one you are importing from).

• Really. Make sure the data field is correct.

• You now imported the classification scheme to another layer. Since this has your highest high value, you can fix the upper class break or breaks. Check that your class scheme is exactly what you want it to be. You can adjust the numbers in the Label column to get rid of excess decimals. This will not affect the classification. Fix the colors so they are what you want. This will be the layer you will use to import the class scheme to all your other layers, including the one you first classified (the column with the lowest low value).

At this point, you should have twelve population change maps, all with the same classification scheme, but different layer names. You have worked through the complexities of data classification, and learned how to use ArcGIS's data classification capabilities.

Color and additional Map Types

To complete this exercise you need to select an appropriate color scheme and also generate a series of graduated, proportional symbol and dot maps with appropriate data and symbolization. Background information on these issues will be presented in lectures and ArcGIS help may also be of help. In each case, make sure you exclude your 999999999 values before mapping, and show your instructor what you are up to.

1) Select appropriate colors for your choropleth maps, but also for the additional map types noted below. Use information covered in the lecture on Color and Maps and you may also consider using the Colorbrewer web site. Please explain in your lab blog why you chose the colors you chose.

HUGE HONKIN' JUMBO TIP: Please do not create seperate data frames in ArcGIS for the additional three map types below. You should use additional data frames if you have several maps with different projections or different areas of the world. Using different data frames when the base map is the same sets off some bug in ArcGIS which can lead to layers being deleted. Put all your map layers - four different map types - within one data frame.

2) Generate a series of graduated symbol maps. Graduated symbol maps require total numbers (not percent change) so you will use the original totals for each of the twelve years of data. Graduated Symbol maps also require your data to be classified. As this is a different set of data from the percent population change, you will have to work through the same process as above to generate a classification scheme that fits each of the twelve sets of data. Please create another twelve layers (in addition to the twelve for your choropleth maps) which represent your total population data as graduated symbol maps. You must pay attention to the various symbolization options (symbol shape, size, color, etc.) you have available to you. Please let your instructor see what you are up to, and justify your symbolization decisions in your lab blog.

3) Generate a series of proportional symbol maps. Proportional symbol maps require total numbers (not percent change) so you will use the original totals for each of the twelve years of data. Proportional symbol maps do not require your data to be classified. Please create another twelve layers which represent your total population data as proportional symbol maps. You must pay attention to the various symbolization options (symbol shape, size, color, etc.) you have available to you. Please let your instructor see what you are up to, and justify your symbolization decisions in your lab blog.

4) Generate a series of dot density maps. Dot density maps require total numbers (not percent change) so you will use the original totals for each of the twelve years of data. Dot density maps do not require your data to be classified. Please create another twelve layers which represent your total population data as dot density maps. You must pay attention to the various symbolization options (dot size, color, etc.) you have available to you. Please let your instructor see what you are up to, and justify your symbolization decisions in your lab blog.

Yeesh! Over forty thirty (still a buttload of) stinkin' maps! Egad.

Next step: getting all of these maps out of ArcGIS - and heading to the web.

E-mail: jbkrygier@owu.edu

...to Geog 353 Main Page and Course Description
...to krygier teaching page.
...to krygier top page.

OWU Home
OWU Geology and Geography Home