My R interface has been pretty basic in the last few years. I have usually stuck to the R console. Yes, I’ve tried Emacs with ESS; a staple, but it is so unbearably antiquated that I always gave up on its significant learning curve. GUI packages–especially Rstudio–offer viable alternatives, but I feel the GUI lets me lose focus of the code. I have been envious of TextMate for Mac, but alas, I’m not a Mac user. Recently, though, I’ve moved to Sublime Text 2. With some nudging, I have been able to mimic the typical R console environment in the more-powerful Sublime Text program.

Want to skip to the basic instructions without the narrative? Just read the block quotes for specific instructions.

Sublime Text 2 offers a similar experience as TextMate (e.g., autocompleting parenthesis and quotation marks) and integrates well with R and other programming languages. R users aren’t the primary target and it takes a little kick to get the environment perfected for the R community. This is a little tutorial to give R users a similar experience with Sublime Text without ruining the program’s versatility. That is, you can have a familiar R experience and easily use Sublime Text with programming languages (e.g., JavaScript, C++) and typesetting programs (e.g., LaTeX, Markdown, MultiMarkdown).

This setup will let you

• Browse R code with syntax highlighting
• Use the familiar R console within Sublime Text
• Send R code from a script to the R control (equivalent to Ctrl + R)
• Organizing the layout so the script and console can be viewed simultaneously.

Initial Setup

I will presume some user knowledge here and will focus on the Windows user. Setup is very easy. Their website has links to executable autoinstallers, you won’t need to unzip and manual install nor are you depending on a third party to graciously compile code on your behalf.

Follow the instructions and install using the default settings.

Take a look around

Before proceeding, take a look at the R interface. Similar to other programs (e.g., Notepad++), Sublime Text offers native syntax highlighting for R code. Load one of your codes or try a sample script. If you don’t see any highlighting, you can always tell Sublime Text by clicking View  > Syntax and choosing R. Otherwise you can look in the lower right-hand corner as is shown in the image.

A nifty feature is autocompleting parenthesis and other special characters. Type “write.csv(” and it’ll auto-insert the right paren “)“. Little underlines help you match each left parenthesis with the right parenthesis. It also works with “{“, “[", single, and double quotation marks.

Comments have a lighter tone. So does the ubiquitous "<-". Functions and special characters also have highlighting. If you don't like the color scheme, you can change it under Preferences > Color Scheme. The default is Monokai, Twilight is a very dark theme while iPlastic is a very white theme.

Once you type a variable or function names, you can easily reference it later for autocompletion. Load the data into a variable called "sampleData" and you only need to type "sa" and hit tab to complete the name. If there are multiple possible matches, then you can use arrow keys to choose. The limitation of this, however, is that Sublime Text treats "sample.data" as two separate names. To counter this, I've started to use underscores, which is friendlier. Also, Sublime Text is using a simple approach--it simply refers to what you've typed before, it is not remembering variables in your memory the same way the GUI interfaces do.

There are plenty of shortcuts. Select a few lines of code and press Ctrl + / to toggle comments. Ctrl + L will select the entire line based on your cursor. Ctrl + G, type in a number and you'll jump to that line of code. Type Ctrl + : and type a variable name, it'll find all instances. If you have multiple tabs, type Ctrl + P and you can jump to specific tabs.

At this point, however, you can't execute or "compile" code. It's just pretty font, helpful quirks, and shortcuts.

Package Manager

Our next goal is to be able use the R console and do calculations within Sublime Text. We're going to use the SublimeREPL package to accomplish this. Don't worry, this isn't a rarefied thing that will clutter your RAM just for R; you'll probably use this package for other things too.

Use Sublime Text's built-in Package Manager. To access it type "Ctrl + Shift + P" to bring up a text box. Without using your mouse, begin to type "Install package" (without quotation marks).

Then type "SublimeREPL" and hit enter when it's highlighted.

SublimeREPL essentially allows Sublime Text to access compilers or other programs. You can write in Sublime Text and send it to R for the computing. The results are sent back to Sublime Text for display. For you, it will operate like the normal R console.

However, we need to tell SublimeREPL where R is located on our computer. We need to add R to the Windows Path.

Find where R is installed. To do this, right click on your R icon--whether it's start menu, desktop, or taskbar. Under the "Shortcut" tab, note the file path under "Target:". Copy and paste this to a text document, you'll need this shortly.

Update: Wojciech Bederski (wuub in comments) gave a great tip to make this even easier than before.

Now,

Click on Preferences -> Package Settings > SublimeREPL > Settings – User. Use the screenshot or example below and paste in the path.

For example, you might paste:

{
"default_extend_env":{"PATH":"{PATH};C:\\Users\\tls573\\Documents\\R\\R-2.14.1\\bin\\x64"}
}

Notice in my example that I am ignoring (disabled) some packages I have installed. In order for this to work, I needed to add a comma after the bracket on line 8 and pasted the “default_extend_env” within the curly brackets.

Now you have told where Sublime Text can look to find R.

Open the R Console by clicking Tools > SublimeREPL > R. A window will pop up with the familiar R console header.

Further Customization: R Shortcuts

The R console is fantastic because you can quickly execute code from a script. SublimeREPL does have this capability, but the R user will want to use Ctrl+R, or something similar. Additionally, the native shortcut to execute script has some issues because it conflicts with other shortcuts already built-in Sublime Text.

Sublime Text calls these shortcuts “keybindings”. Fortunately, they can be edited. However, you have to be careful since Sublime Text has a lot of shortcuts and it’s easy to duplicate them, causing further issues.

I have a proposed set that should be pretty R user friendly. You can select some code (multiple, partial, or single lines) and run it in R using Ctrl + Shift + R. This is similar to the shortcut in the R console (a previous conflict prevents Ctrl + R without a lot of work).

Find the existing keybindings for SublimeREPL. Navigate to “C:\Users\[login user]\AppData\Roaming\Sublime Text 2\Packages\SublimeREPL\Default (Windows).sublime-keymap”. Right click on the file and open it in Sublime Text. Where it says [login user], use your Windows login without the square brackets.

Copy and paste the contents of the code below into the file, completely replacing the text, then save using the same file name.

Wuub (Wojciech) pointed out that the default keymap will be overwritten with every REPL update. Use this method instead:

Go to Preferences > Key Bindings – User and paste the code below. If something if already there, then append the code to the bottom.

You can also execute a line of code–without highlighting–using Ctrl + Alt + R. Select line(s) of code and send it to the console by typing Ctrl + Shift + R and then hitting R again. Similarly, display a line of code using Ctrl + Alt + R and then typing R again.

If you want to execute the entire script, type Shift + F7 and the entire file will be ran in R, no matter where your cursor is placed. I chose F7 because that is commonly used to compile code in other languages in Sublime Text.

A block of code can be sent by pressing Ctrl + Alt + Shift + R.  A block is the code between { }, such as a function. If you want to view the code before it’s executed, press Ctrl + Alt + Shift + R, then press R again.

Side-by-Side

It is difficult when the script is in one window and the console in another, but it’s easy to have the same feel as the R console.

To have the script and console side-by-side, click View > Layout. Choose either “Columns: 2″ or “Rows: 2″, depending on your preference.

Switch between the side-by-side windows using Ctrl + 1 or Ctrl + 2.

Now you can see your script(s) and output side-by-side.

Drawbacks

The annoyances:

• The console behaves like a text editor. Your cursor will need to be at the end of the line before you hit enter. Otherwise it just creates a linebreak.
• When you use Ctrl + Shift + R (send selection) or Ctrl + Alt + R (send line), the console shows “>” every time a command is send. This is a little annoying, but moreso if there are errors in your code.
• Sometimes the Ctrl + Shift + R or Ctrl + Alt + R is a little finicky, but I think I do it too quickly. Be deliberate if it does not seem to work.
• When I am doing some heavy R computations, the setup can become slow. It has only happened once and could have ultimately been my fault. Let me know about any performance issues.

Besides R, Sublime Text can also handle other programming/scripting language. It can also handle typesetting languages, such as LaTeX or Markdown. SublimeREPL also works with other languages, such as Ruby and Python.

Credits

This is a tutorial that combines the work of others. This tutorial is just making gumbo from the recipes of others. In particular, Wojciech Bederski, who wrote the SublimeREPL package, which is key. Update: Wojciech also wanted to give props to hootener, who really supported the Windows R implementation for REPL.  Mads Hartmann Jensen wrote added the R support in REPL.  Obviously, the Sublime folks have put together a…uh…sublime program.

I agree with the thesis, but not the logic. Dr. Rhoads attempts to tie the recent enrollment declines to limited budgets. “Enrollment in community colleges across the country is plateauing and declining,” he argues, “despite rising student demand.” Dr. Rhoads seems insistent on this latter part, but here is where I get lost. Although demand is much higher than 5 years ago, many states have had declining enrollments since two years ago.

Moreover, the sharp decline in enrollment was projected ahead of time based on demand-based factors*. I predicted the sharp decline in Iowa’s community college enrollment because of decrease in demand.* The U.S. Department of Education estimated enrollment would move from a 7% growth between 2008 and 2009 to a 0.8% growth between 2009 and 2010 (turned out to be a 1.7% growth) using a methodology primarily based on demand. Oddly, Dr. Rhoads report also cites a Community College Review article that notes several states expected their enrollments to decline, even in states where he claims students are being turned away because of budget cuts.

The reason for the decline? I think that many students who jumped into community colleges simply graduated, transferred, or entered back into the labor market. Dr. Rhoads, during a bit of a Twitter conversation, noted that the graduation rates have not increased at community colleges, and he’s absolutely right. But the graduation rate formula at community colleges are woefully inadequately, capturing first-time, full-time students. In this instance, it’s more meaningful to look at total number of awards. In Iowa, we saw the largest jump in the number of awards since we started tracking that statistic. Unfortunately, I haven’t been able to grab the most recent national statistic.

It is also likely that there was a reduction in student demand as unemployment began to recover. Although the unemployment rate has been slowly declining, there has been a reducing in the number of unemployed individuals. In August 2009 (the fall 2009 semester), there were 14,823,000 unemployed “adults” (16+). At the beginning of the fall 2011 semester, there were 14,008,000 unemployment adults. Currently, as of February 2012 (let’s call that the spring 2012 semester), 13,430,000 unemployed adults.

The Converse

Like I said, I ultimately agree with Dr. Rhoads. The budget cuts, “operating more with less”, has taken a toll, but not in the way he suggests. Instead, I suspect community college enrollment in specific majors/programs was artificially capped during the enrollment surge. Enrollment in community colleges grew 7% in 2009, which was likely caused by the sudden economic deterioration. During this point, we saw colleges struggle to add capacity to technical programs which could not expand fast enough to keep pace. Within Iowa, there were numerous instances of programs with long waiting lists.** At this point, Dr. Rhoads is right; many students were turned away and, likely, those who would benefit the most from college.

Rejoinder

Dr. Rhoads is a respected scholar and the report seemed to catch the attention of other good education researchers. The line of argumentation doesn’t make sense though. Student demand is clearly higher now than 5 years ago, but there is no way enrollments are going to be as high as 2009. Yet, the report compares current enrollment with the enrollment peak as evidence that students are being turned away.  California is a clear case where this is happening, but the evidence isn’t clear in the other states***

These seem like minor quibbles. It would be easy agree to disagree with the logic for an argument I ultimately agree with. But I think it is very important for academia to play a mediating role between government, administrators, and the public. Lawmakers who are bent on reducing public expenditures will claim that higher education is well-funded or even over-funded–even when they are clearly not adequately funded, such as the case now. Higher education administrators will complain they are under-funded or have too many burdens even in periods where funding is adequate and the regulation is meaningful.

And some of it is my past experience. I hated, hated, when an advocacy paper from a think-tank was released that played fast-and-loose. As a result, there was a lot of running around trying to adjust the newly-formed misconceptions. Normally, these reports argued against my belief or policies. In this circumstance, there is a peaceful rejoinder on an unfortunate circumstance–we ultimately agree that students are being turned away, just differ on when it happened. Even when community colleges had rapidly increasing enrollment, the limit on programs seemed to conflict with their open admission policies. Although students were admitted and were suitable for some programs, there was simply no room.

It will mostly include graphs and data visualizations, but will also use other sources–such as art and movies–for ideas, such as color palette. Essentially, I hope it will serve as a sort of free style magazine for those starting new projects. I’m also accepting submissions of any interactive, static graphing style, data visualization project, or anything that might help someone choose a color palette, graphing style, or layout for their own project.

My new publication in November’s IR Applications, which is published by the Association for Institutional Research:

This paper explores the relationship between student major and industry of employment and its application to higher education accountability. Data provided by statewide longitudinal data systems (SLDS) have enabled state educational agencies and colleges to follow students into the workforce. While most studies have focused on wage outcomes, this study shows how to use SLDS data to understand the correlation between major and industry. The transition into the workforce is an important outcome since it is an assessment of a college’s ability to develop specific, targeted sectors of the economy. We use SLDS data from Iowa to follow community college alumni from 2002 through 2008.

Find it here.

Darker areas represent a higher proportion of kids in poverty. Pockets of high poverty, 30%+, are predominately in southern Iowa. Not surprisingly, educational attainment for adults in these same school districts also tend to be lower.

The lower-left graphic–dependence on manufacturing jobs–is one of the first items I mention to folks when discussing Iowa’s economy. Rural areas rely more on manufacturing than agriculture as a percentage of employment and as a percentage of income after the 1982 recession–most assume it’s agriculture. The Federal Reserve Bank of Chicago noted the greater reliance of non-metro areas on manufacturing in the past decades. In the graph below, the orange areas represent counties which the share of manufacturing income is greater than the statewide average. The white areas are metropolitan areas. In 2009, two-thirds of non-metro counties are highlighted in orange.

Scholars have noted the decline in rural Midwestern states. Often, they point to the high concentration of manufacturing as a perilous position for rural counties. Manufacturing, one could argue, is often outsourced, which would be catastrophic for these communities if a manufacturer left. It is critical to note, however, manufacturing in these areas are often forms of food manufacturing. For instance, corn is rarely sold for corn, but is now used to create products such as high fructose corn syrup and, of course, ethanol (40% of all corn production). Though ethanol is not considered food manufacturing, it is similar insofar as it is the processing of sugars. Manufacturing provides greater gross value-added than the agriculture cultivation process which contributes to a greater proportion of income coming from manufacturing as opposed to agriculture. In reality, manufacturing is still tied to Iowa’s agricultural base, making it harder for those firms to be simply exported.

Iowa has had increasing reliance on food manufacturing for employment. A graph from the aforementioned Federal Reserve Bank of Chicago shows the increase in Iowa in relation to neighboring states:

The rise of manufacturing is notable, but it is a mistake to separate it from Iowa’s agriculture. Fortunately, it means rural Iowa is not just waiting to be outsourced, but it doesn’t guarantee economic growth either. Notable books and articles have touched on the “death” of the rural Midwest–search for any book with “Middle” in its title–which is mostly tied to the shift “away” from agriculture and large migration outflow.

Iowa, where research in Hollowing Out the Middle was based, has transformed from agriculture as a final good to agriculture as an input into manufacturing. The greater value-added from manufacturing–such a Marxian concept, I know–has helped keep real income higher in the non-metro areas.

This data does not negate the other concerns raised by the authors of the aforementioned books, but non-metro areas can provide a base for manufacturing that may not be easily outsourced.

There is a relative floor at $40,000 with a wide variation of unemployment (poor clinical psychology). There is a negative correlation between earnings and unemployment rate, but it might be too presumptuous to presume that high-paying majors get their cake and eat it too. Depending on the methodology, the higher earnings might simply reflect the fact they have a job.

Of course, median earnings is only one dimension. Below shows a messy graph that also incorporates the first and third quartiles:

On average, the third quartile was about 44 percent above the median values. Most of the majors had median earnings in the center of the inter-quartile range.

Similar data is also available for Iowa community colleges.

There are important threads in the data with are important to notice. The top 10 paying jobs are:

1. Petroleum engineering
2. Pharmacy/Pharmaceutical sciences & administration
3. Mining and mineral engineering
4. Naval architecture & marine engineering
5. Nuclear engineering
6. Mathematics and computer science
7. Chemical engineering
8. Electrical engineering
9. Metallurgical engineering
10. Military technologies

H1-B visa applications contain the company name, the occupation they’re attempting to hire, the “target” salary for the worker, and prevailing salary taken from the Census or Bureau of Labor Services. Each visa application is a public record which you can use to search salaries for a new job or promotion. Immihelp is one site where you can search a database of H1-B applications, below is a screenshot for a Senior IT Application Analyst at Principal Life Insurance

If the same occupation as you’re applying for is listed, then you should at least aim for the target salary. The prevailing wage amount gives you a sense of the minimum acceptable wage (unless you’re demanding more) since it’s the typical wage for that particular job in the U.S. In the above example, the prevailing wage data was taken from the Occupational Employment Statistics survey.

There are a few sites which attempt to provide the same service, such as GlassDoor and PayScale. These are limited since they rely on self-reported data and seldom contain all occupations. The H1-B visa database has it’s own limitations since it only contains data from applications. This is especially applicable if you’re in a field that doesn’t require a Bachelor’s degree or your company hasn’t filed for a visa application in recent years.