Live Regression

Entering your data

To enter your own data, you must

• Prepare the data set
• Put it on the web using Dropbox
  
• Shorten the URL using bit.ly
• Enter the shortened URL into the widget
  

Once the data set has its short bit.ly URL, then anyone can use the widget on your data using the URL.

Preparing the data set

The data set must be a text file in CSV (comma separated values) format.  In other words, it should look like this:

GDP,Population,GeographicSize,LatitudeofCapital,LongitudeofCapital
10380.38,1372.45,9573,40,116
2049.50,1278.11,3166,29,77
17418.93,321.961,9526,40,-77
888.65,255.780,1905,-6,107
2353.03,204.998,8516,-16,-48
250.14,190.595,882,34,73
573.65,182.202,924,9,7
185.42,159.140,148,24,90
1857.46,146.605,17098,56,38
4616.34,126.832,378,36,140
1282.73,121.005,1964,19,-99
284.93,102.152,300,15,121

The first line of the text file is the list of variable names, and each line after that contains numerical values separated by commas.  The data above is for the largest twelve countries by population (China, India, United States, Indonesia, Brazil, Pakistan, Nigeria, Bangladesh, Russia, Japan, Mexico, and the Philippines).  Variable names should be alphanumeric (no special characters), and there should be no missing data entries.

Excel and most statistical software can export your data as a CSV file. 

Put your data set on the web

The easiest way to put your data set file on the web is to use Dropbox.  If you use Dropbox, make sure your file is in your Dropbox folder.  Then log into dropbox.com, find your file in the file picker, in this case countries.csv, and then click the "Share" button which appears on the right.



Dropbox will respond by giving you the following dialog box which includes a URL which is your "Link to file".



Take this URL

https://www.dropbox.com/s/omlqcl47k9cf0c4/countries.csv?dl=0

and now all you have to do is to delete the dl=0 part at the end and replace it with raw=1 like this:

https://www.dropbox.com/s/omlqcl47k9cf0c4/countries.csv?raw=1

Shorten the URL

Since this URL is too long to put into the widget (the widget is written in Mathematica, which disallows text inputs in widgets), next you shorten it using bit.ly.

Go to bit.ly and then paste the URL into the box at the top of the page, and then you will receive a shortened link such as this:

http://bit.ly/1OgRUN1

Once you have created this URL, it is "permanent" and you don't have to go through all the steps above just to use the widget again.  If you revise the data file on your local computer, the widget will import the revised data when you click the "Import Data" button in the widget.  If you move the file to a different location on your computer, however, the link will no longer work and you will have to create a new one.

Using the widget

Now you can enter this short link into the widget and press the "Import Data" button at the top left.

If you just want to try out the widget without importing your own data, just click on the "Example data sets" button and you can choose among some examples there.

Choose the dependent variable from the dropdown menu.  In the example below, the dependent variable chosen is GDP.



Here GDP is regressed on three specifications.  In the first, GDP is regressed on Population.  In the second, GDP is regressed on Population and GeographicSize.  In the third, GDP is regressed on LatitudeofCapital and LongitudeofCapital.  A constant term is included in all of these regressions (right now there is no way to turn off the option of including a constant term).  To change the number of specifications (up to a maximum of 10), use the drop-down menu.

To include or not include an independent variable in a specification, simply click on the variable name in the row of buttons.  If the set of independent variables is empty or includes the dependent variable, an error results and no regression results are shown.

Another way to include an independent variable in a specification is to command-click (or alt-click, if you are using a Windows computer) inside the output window.  In the example below, the command-clicks (the blue squares) cause LatitudeofCapital and LongitudeofCapital to be included in the first specification.



To clear the command-clicks, press the button "Clear command-clicks."

If you click the arrow in front of "options," various options will appear. 



The first pop-up menu allows you to display either standard errors, t statistics, p values, or 95% confidence intervals of the estimated coefficients (or nothing at all).  These are displayed in parentheses and in gray underneath each estimated coefficient. 

If you check the stars checkbox, then stars will appear next to the coefficients.  Three stars indicates the coefficient is significant with probability 0.999 or greater.  Two stars indicate it is significant with probability 0.99 or greater.  One star indicates that it is significant with probability 0.95 or greater. 

You can also choose how many significant digits are shown for the fractional parts of numbers.

If you click the arrow in front of "appearance," you can adjust the appearance of the output.



If you check the "shrinkblankrows" box, then rows of white space are deleted; this is useful if there are many possible dependent variables but your specifications only include a few of them.  You can choose the font size (in points) of the output font (Times).  Finally, you can use the spacing slider to change the horizontal spacing between specifications and the vertical spacing between rows.

Please explore all of these controls.  Please send any comments or suggestions to michael@chwe.net.