Search
   >> Home >> Resources & support >> FAQs >> Working with tmap and maps
Note: This FAQ is relevant for users of releases prior to Stata 10.

How do I graph data onto a map with tmap?

Title   Working with tmap and maps
Author Kevin Crow and William Gould, StataCorp
Date April 2006; minor revisions May 2007

Introduction

With tmap, you can graph data onto maps and produce results such as these.

tmap is a user-written command by Maurizio Pisati. This FAQ explains how to use tmap. The process is as follows:

  1. Obtain and install the tmap, shp2dta, and mif2dta commands.
  2. Search the web for the files that describe the map onto which you want to graph your data. You can use ESRI shapefiles or MapInfo Interchange Format.

    ESRI shapefiles are the more common. In this format, there are three files associated with a map: an .shp shape file, a .dbf dBASE file, and an .shx index file. You only need the .shp and .dbf files. You translate those files into a format usable by Stata with the shp2dta command. Doing so creates two .dta datasets, one corresponding to each file.

    The MapInfo Interchange Format consists of two files with suffixes .mif and .mid. You translate these files with the mif2dta command. Just as with shp2dta, doing so creates two .dta datasets.
  3. Look at the translated .dbf (.mid) file. It is a .dta dataset and you just use it. Examine the dataset to determine the coding used by the map's authors to designate areas. For instance, 1 might mean Alaska and 2 Alabama in one dataset, and 1 might mean Albania and 2 Argentina in another.
  4. You have data you want to plot onto a map. Let’s assume that the data are stored in a Stata .dta dataset. You need to modify your .dta dataset to use the same location coding as that used in the map. Call that variable id.
  5. Merge on id the translated .dbf (.mid) dataset with your dataset containing the statistics to be graphed.
  6. With the merged dataset in memory, make the graph by using tmap. You will tell tmap about the other translated dataset (the coordinate dataset) by using an option.

Step 1: Obtain and install the tmap, shp2dta, and mif2dta commands

Type

 . ssc install tmap

 . ssc install shp2dta

 . ssc install mif2dta

You need to perform this step only once.

Step 2: Find a map (an ESRI shapefile or a MapInfo Interchange Format file)

A map records the geometry and attribute information of spatial features. Those maps are available from public and private sources. You can use maps recorded in either of two formats:

  1. ESRI shapefiles. This format was developed by the Environmental Systems Research Institute, Inc. Shapefiles come in several types. You will want a "polygon shapefile", the form suitable for maps. The map is stored in three files:
    .shp, the coordinates;
    .shx, an index; and
    .dbf, the codings.
  2. MapInfo Interchange Format. This format was developed for use with the MapInfo software. The map is stored in two files:
    .mif, the coordinates; and
    .mid, the codings.

It is usually easier to find ESRI shapefiles than MapInfo Interchange Format files, but you may use either.

Say you want to find a map of the United States. Using a search engine such as Google or Yahoo!, search for "United States shapefile". One result is described as "This dataset is a polygon shapefile containing the states and territories of the United States ...". We found http://www.nws.noaa.gov/geodata/catalog/national/html/us_state.htm and clicked "Download Compressed Shapefile". We unzipped s_14jl05.zip, which contained the following files:

s_14jl05.shp
s_14jl05.shx
s_14jl05.dbf
These are the filenames as of May 2007. They will most likely change over time.

We need only two of the files, s_14jl05.shp and s_14jl05.dbf.

Had we searched for a MapInfo map, there would have been only two files, and they probably would have been called s_14jl05.mif and s_14jl05.mid.

Step 3: Translate the files

With the files we just extracted in the current directory, in Stata, we type,

 . shp2dta using s_14jl05, database(usdb) coordinates(uscoord) genid(id)

Pay attention to the three options we specified:

shp2dta can take several minutes to run, depending on the map's size and level of detail. The U.S. map, however, took only a few seconds.

We would have translated MapInfo files the same way, but we would have used the command mif2dta instead of shp2dta.

In any case, the translation has created two new .dta datasets: usdb.dta and uscoord.dta.

Step 4: Determine the coding used by the map

To determine the coding used by the map's authors, type

 . use usdb, clear

 . describe

 Contains data from usdb.dta
   obs:            56
  vars:             6                          29 Mar 2006 11:52
  size:         2,744 (99.9% of memory free)
 -------------------------------------------------------------------------------
               storage  display     value
 variable name   type   format      label      variable label
 -------------------------------------------------------------------------------
 STATE           str2   %9s
 NAME            str24  %24s
 FIPS            str2   %9s
 LON             double %10.0g
 LAT             double %10.0g
 id              byte   %9.0g
 -------------------------------------------------------------------------------
 Sorted by:  id

  . list id NAME in 1/5
 
      +-------------------------------+
      | id                       NAME |
      |-------------------------------|
   1. |  1   District of Columbia     |
   2. |  2   Arizona                  |
   3. |  3   Ohio                     |
   4. |  4   California               |
   5. |  5   Alabama                  |
      +-------------------------------+

Let’s shift away for a minute from the details of this map and talk about the graph we want to draw. We want to graph population by state, and we have a dataset named stats.dta containing population figures. In our dataset, we have states recorded using a different coding, and the identification variable is called scode.

We must modify our dataset to use the same coding as the map, and the variable containing the codes must be named id.

To achieve our goal, we made an intermediate dataset called trans.dta that contained two variables, scode and id. Each observation records equivalent codes. When we created trans.dta, we happened to look more carefully at usdb.dta. We discovered that the map dataset contained information about not only U.S. states, but also territories. We will just ignore that extra information. Our trans.dta dataset records only the 51 observations we care about, one for each state plus Washington, D.C.

Then we merged our stats.dta with trans.dta based on scode:

 . use stats

 . merge scode using trans, sort unique

To ensure that there were no errors, we checked that all observations matched (_merge==3) and then dropped the _merge variable:

 . tabulate _merge
   (output omitted)

 . drop _merge

Step 5: Merge datasets

We now must merge stats.dta with usdb.dta from the map, and this merge is based on the id variable:

 . merge id using usdb, sort unique

Because our map includes locations not included in our original data, namely, territories as well as states, there will be observations in usdb.dta that are not also in stats.dta. We should check our merge:

 . tabulate _merge

Here we expect all _merge values to be 2 and 3. If our map did not include territories, or if our original data did, we would expect all _merge values to be 3.

Finally, drop the unnecessary observations:

 . drop if _merge!=3

Step 6: Draw the graph

To draw the graph, type

 . tmap choropleth pop1990, id(id) map(uscoord.dta) palette(Blues)

We will soon deal with Alaska and Hawaii and the effect they have on our graph. Right now, focus on what we typed:

 . tmap choropleth pop1990, id(id) map(uscoord.dta) palette(Blues)

Choropleth is not the name of a variable in our dataset; it is the kind of graph we want to draw. In a choropleth graph, different areas have different colors. tmap can draw other kinds of graphs, too.

Let’s go over the options we specified:

In the command

 . tmap choropleth pop1990, id(id) map(uscoord.dta) palette(Blues)

we specified variable pop1990, and in the dataset, that variable contains the population. The units do not matter; the data could just as well be coded in millions and we would have obtained the same graph, although the legend would change.

By default, tmap choropleth divides the specified variable into four groups that are based on quartiles. You can change the number of groups by using option clnumber(#), where # can be between 2 and 9.

We will stick with four groups. However, we want to exclude Alaska and Hawaii from our graph. To do that, type

 . tmap choropleth pop1990 if id!=13 & id!=56, id(id) map(uscoord.dta) palette(Blues)

or

 . tmap choropleth pop1990 if NAME!="Alaska" & NAME!="Hawaii", id(id) map(uscoord.dta) palette(Blues)

because 56 and 13 are the id codes for Alaska and Hawaii, and because our dataset happens to contain variable NAME, which records the name in string form, we obtain this graph:

Look closely at the legend and you will see that the population ranges are displayed in scientific notation. You can change the display format with option legformat(format). You might specify legformat(%20.0f). Or you can change the units of the variable. We will change population to be recorded in millions:

 . replace pop1990 = pop1990/1e+6

The legend is also too small. You can make the legend bigger with option legsize(#), where # specifies a text-size multiplier, such as 2. Our improved graph is shown below:

 . tmap choropleth pop1990 if id!=13 & id!=56, id(id) map(uscoord.dta) palette(Blues) legsize(2)

tmap has many other options. Read about them in the online help file (type help tmap) or in the original article by Maurizio Pisati (2004).

Friedrich Huebler&rquo;s blog, at http://huebler.blogspot.com, occasionally discusses tmap.

Reference

Pisati, M. 2004. Simple thematic mapping.
Stata Journal 4: 361–378.
The Stata Blog: Not Elsewhere Classified Find us on Facebook Follow us on Twitter LinkedIn Google+ Watch us on YouTube