Summary Function that is compatible with xtable

If you like to make nice looking documents using Latex, I highly recommend using the 'xtable' package. In most instances, it works quite well for producing a reasonable looking table from an R object. I however recently wanted a LaTeX table from the 'summary' function in base R. So naturally I tried:

Which gave me the following error:

Error in xtable.table(summary(foo)) :
  xtable.table is not implemented for tables of > 2 dimensions

So I decided to create a simple function that would return a dataframe which is easy to use with xtable. Here is what I came up with.


Now, when I try to use xtable I get the following output:

This should lead to an easier way to incorporate more summaries when you are writing your paper, using R and Knitr of course. If you do use knitr, make sure to try the results = 'asis' option with xtable from R.

Leave a Comment

Filed under R

Sending Email From R

When I am am working in Sql Server and need to send an email I use "sp_send_dbmail". So when I am working in R, I didn't know how to send an email. I often use this as notification that a process has finished. It also works nicely as a text to your cell phone. I had one additional reason why I wanted to be able to email from R. I wanted to send an email to my evernote account with just a few key strokes. The goal was to accomplish this by writing a simple wrapper function. Below is the solution that I came up with. It works, but there are serious security implications. I offer this merely as a proof of concept. Hopefully someone can show me a better way to handle passing your email password to the R function.

I often have an R terminal open, so when I have a great idea for a research project I can add a note to my evernote account simply. For instance:

en(s="Read up on current imputation methods",tags="#Research")

Then a new entry is added for me in evernote with the tag 'Research'. ( I have noticed that the tagging seems to only work if the tag previously exists in my evernote account.)  Often I have a task that just needs done that I don't want to forget about. I can issue a quick command and then I will have a record of it.

en(s="email adviser about research")

That is all I need to do and the note is added to my account. I have found this to be quite useful and hopefully you will as well.

Leave a Comment

Filed under email, R

Job Market For Statisticians

I have been forced to think about the job market lately. It started with a class assignment which was meant to simply open my eyes to current job market. I felt that I was already familiar enough but completed the assignment to be a good student. I completed the assignment and outlined the skills I need improve upon and so forth. With in day of completing my assignment I came across "A Guide and Advice for Economists on the U.S. Junior Academic Job Market: 2014-2015 Edition" after clicking through some links on facebook. I found it a great read and it caused me to starting thinking about a few things that will likely prove helpful down the road. I had originally intended to pursue a Ph.D. in Economics after finishing my M.S. in Statistics. However, life took a turn and I ended up working full time and then started working on my Ph.D. in Statistics part time while continuing to work. Make sure that you take a look at the salary tables that are included. The table below is for full time working White Males by which Ph.D. they obtained. There is more variation associated with the Economics degree, but not enough to not make it look better than Math or Statistics based solely upon salary.

For White Males Median Salary SE       95 % Range
Mathematics/Statistics  $100,000   1,500  (97,000 - 103,000)
Economics   $126,000   5,500 (115,000 -137,000)

(Data taken from: Table 50 , the 95% range is mine based on the assumption of a normal distribution.)

There is also an article in the American Statistician recently about career paths, "Which Career Path Will You Follow?". Between these three events that occurred within a week, I thought that it merited a post. Have some other useful job advice or interesting statistics that current graduate students should know? Post a comment.

Leave a Comment

Filed under Uncategorized

Helper Stored Procedures in SQL Server - Part 2

Continuing with last week's post about Helper Stored Procedures, I wanted to introduce two more that I use frequently. The first one is a built in stored procedure (  using SQL Server) called 'sp_helptext'. This stored procedure allows you to view the content of other stored procedures (SP) and more, here is a link to the Microsoft documentation. Here is an example of how you use it:

This will show you the contents of sp_helptext  by using it. This comes in very handy as a keyboard shortcut. I assigned mine to 'Ctrl+3'.  Often I have a SP that does data manipulation and puts the final data in a table that occurs as the last step of my SP. I may not remember which table the data goes into, but I can simply highlight the SP ( I typically double click) and then use 'Ctrl+3' and the contents of the SP appears in my results tab. This also works nicely for function and triggers that you might use.

Another piece of information that I am often curious about is how large is a particular table. If you like going through menus, then you can left click on a table name in the object explorer and choose properties. Then select 'storage' from the left navigation menu. You will see under the 'general' section: "Data Space","Row Count" and "Index Space". While this is okay if you only perform this task once a month, it takes way to long if you are running this 5 + times per day. I wrote a stored procedure called 'sp_tablesize' that is a wrapper for another Microsoft SP, that is sp_spaceused.

After you create the SP 'tablesize_SP' you can assign it to a shortcut in SSMS. I choose to use 'Ctrl+4' for myself. We can create a table and add data and an index to make sure that our new SP works.


You will notice that our index size increased after adding one, which is to be expected. Hopefully this will allow you to quickly check the size of a table. Combine that with 'sp_helptext' and 'sp_top10' you can find out a lot of information about a table quite quickly. Let me leave you with a quick reference guide of which SP to use depending on what you want to know.


SP Name Use Shortcut
sp_helptext View the contents of a stored procedure or function Ctrl+3
tablesize_SP Check the size of a table including indicies Ctrl+4
sp_top10 View the top 10 records of a table Ctrl+1
sp_help View Information about an object for instance table schema Alt+F1

Note: SSMS 2012 was used for this example.

Leave a Comment

Filed under Sql Server

Helper Stored Procedures in SQL Server - Part 1

Often I have days where a large amount of my time is spent writing SQL code in SQL Server Management Studio (SSMS). Building tables, creating indices and verifying data are all common tasks. However, I often myself needing to checking which columns are in a table or to take a quick look at the data. I got tired of typing Select Top 10 * From [MyTable] more than five times per day.   In order to get around typing that repeatedly I made a function called 'sp_top10'.

This stored procedure can be invoked by typing

exec sp_top10'YourTableHere'  where you substitute in your own table name. The real power comes when you assign your new stored procedure to a shortcut. In SSMS you can do this by going to Tools > Options > Keyboard > Query Shortcuts .  Simply add  'sp_top10' to an open key combination and then restart SSMS. You should now be able to highlight ( or doubleclick) a tablename and then use your new shortcut to view the top 10 records in that table. I set my shortcut to 'Ctrl+F1'.

An example:

You output should look similar to figure 1.


Figure 1.

Feel free to edit the SP to suit your particular situation. I use this in conjunction with 'sp_help' which has a default shortcut of 'Alt+F1'  in SSMS almost everyday. So I can select a tablename and 'Alt+F1' if I need to see what type of columns I have and then 'Ctrl+1' to view the first ten records of that table. Hopefully this will help save you time in your daily work.

Note: SSMS 2012 was used for this example.

1 Comment

Filed under Sql Server

Making a Code Book in Sql Server

While working through a coursera course recently ( I started to think about how a code book could be implemented in sql server. There are 3 points that Jeff Leek makes regarding a code book and they are as follows:

  1. Information about the variables ...
  2. Information about the Summary Choices  you made
  3. Information about the study design you used

My focus for this article will be able point number one, information about the variables. Often times it would be nice to know more about your data than simply what data structure it is defined by. For a simple working example, load the Iris datasets into Sql Server and then add some metadata about the columns to provide further information for the users.


Now, you should have a table called Iris with some basic metadata about the columns. While you can navigate to the extended properties through the object explorer, it would be nice to access this information through a query.

Now you have the framework for creating a data code book that can be self contained within the table itself. This will prove most useful when you can can share a sql server table with someone else.


Leave a Comment

Filed under Sql Server

Unexpected behavior with summary function in R

I often find myself working with data that includes dates and times. Sometimes I am interested in looking at what happened on on a particular calendar day. I usually avoid dealing with actual datetime formats when working at the day level and prefer to use an integer for representing a day. For instance the first day of each quarter can be represented as integers (numeric also works for this example). For instance if I wanted to know the oldest date in my dataset I can just take the minimum since I am using a numerical data structure.

[1] 20140101

For some reason which I don't recall, I tried using the summary() function on my dates. The only values that would be valid are the minimum and the maximum, or so I thought.

The output:
# Min.            1st Qu.       Median      Mean          3rd Qu.       Max.
# 20140000 20140000 20140000 20140000 20140000 20140000

This behavior which seemed odd to me, is caused by the way summary() deals with numerical data. So I decided to look at what summary is actually doing. To view the code behind the summary function type:

The portion of code that we are interested in is this:

This code chunk gets executed when the object that we pass to summary is numeric.  If we substitute in our object 'mydates' we get the following code.

If you step through the code line by line, you will notice that after line 4, summary produces what we would expect to see for a min and max value. However, after you execute line 5, the numbers are changed because they are not the actual numbers but they are changed to be significant figures. For example try:

[1] 20140000

So be careful when using generic functions if you don't know what they are doing. I would encourage to take a look at the code behind some of the R functions you use the most. For instance using the fivenum() function does not change my min and max values the same way summary did.

Leave a Comment

Filed under R

Building a productivity system in R, Part 1

I recently came to the conclusion that I need a more meaningful way to track my productivity than the spreadsheet I am currently using, so my next few posts are going to be about building a system in R to track this.  If you're building your own productivity tracking system then by all means take this as inspiration, but don't expect it to suit your needs.  I'm making it to suit my needs using terminology that is common in my workplace and you'll have to figure out what will work for your needs in your workplace.

As with all such endeavors, the thing that is really going to make or break this tracking is the data model, so let's define that first.

At the very top level I have projects.  Each client will have one or more projects.  I'm not interested in tracking work for particular clients (at least for now) so I'm skipping that level, but it is necessary to note that each client has a 4 digit number.  Each project also has a 4 digit number, so the combination of the client digits and the project digits form a partial billing code.  The addition of the task-level 4 digit number makes a complete billing code that can be entered into my timesheet, but we're not there yet.  At the project level, the first two quartets is all that is necessary.  Additionally, we're going to have a name for the project, the date the project gets added, and the date the project gets removed.  Projects can often be multi-year endeavors, so understanding just how long you've been working on various tasks for a project can be useful.  For referencing across different datasets in this data model a project ID will also be defined.

Below the project level, as mentioned, are tasks.  Each task is a concrete goal that has been assigned for me to work on for that project.  Sometimes I only have one task for an entire project, other times I might have several tasks simultaneously. Some tasks may also depend on the completion of other tasks.   So we're going to want the following things: task ID, task name, project ID, complete 12 digit billing code, if the task depends on the completion of another task, add date, complete date, budgeted hours, total used hours (will be cumulative), impact, effort, and notes.  I'm using the impact and effort fields to automatically assign priorities.  They will each be given a value from 1 to 10, with 10 being the highest.  I'm not going to get into how impact and effort will be used to create the priority since I will go into more detail about that in a future post, but see this article for my inspiration.

Finally, I want to track the actual hours in the day that I do the work.  So for this dataset I just want the task ID, the date/time in, and the date/time out.

Since I want all of this to appear as a single object I'm going to use a list containing three data frames.  Below is a function that will actually generate this object.  I expect I'll only ever have to use it once, but it's still useful to me to think in this way.  My next post will get into adding projects and tasks.


Leave a Comment

Filed under Functional Programming, Productivity, R, Uncategorized

Assumption Checking - Part I

Often when working, we are under deadlines to produce results in a reasonable timeframe. Sometimes an analyst may not check his assumptions if he is under a tight deadline. A simple example to illustrate this would be a one sample t-test. You might need to test your sample to see if the mean is different from a specific number. One assumption of a t-test that is often overlooked, is that the sample needs be drawn randomly from the population and the population is suppose to follow a Gaussian distribution. When is the last time in the workplace that you heard of someone performing a normality test before running a t-test? It is considered an extra step that is not usually taken. It should really not be considered a burden and can easily be accomplished with a wrapper function in R.

We can combine that with another function to produce a density plot.

Now, let's see how our functions work.  If we generate some random values from a Gaussian distribution, we would expect it to "normally" pass a normality test and a t-test to be performed. However, if we had data that was generated from another distribution that is not 'normal', than typically we would expect to see the results from the Wilcox test.

Density Plots

Results from 'mytest(normal)':

One Sample t-test
data: x
t = 0.5143, df = 999, p-value = 0.6072
alternative hypothesis: true mean is not equal to 0
95 percent confidence interval:
-0.04541145 0.07766719
sample estimates:
mean of x

Results from 'mytest(chisq,value=5)':

Wilcoxon signed rank test with continuity correction

data: x
V = 214385, p-value = 8.644e-05
alternative hypothesis: true location is not equal to 5

The benefit of working ahead can be seen. Once you have these functions written you can add them to your personal R package that you host on github. Then you will be able to use them whenever you have an internet connection and the whole R community has the chance to benefit. Also, it is easy to combine these two functions into one.



Leave a Comment

Filed under Assumption Checking, R

Read a bunch of csv's quickly

Let's say you have a whole lot of csv files in your working directory.  By some convenient act of divine grace they also happen to have the same column structures.  Reading them into R can be a slow matter as a new R user may try to write out the name of every file, assign it to a variable and then rbind() it all together later on.  A slightly more experienced user might choose to automate it a bit by using list.files() with a for loop to iterate through every csv file in the directory.  A yet more advanced user could figure out via much cursing and pain how to do this using the apply() family of functions, which may actually be the quickest way to do this.  For myself, I like to take headache-saving shortcuts when possible and still maintain some semblance of code efficiency, so naturally I use the eponymous plyr package for this task.

Just for example, let's make a bunch of fake csv files that will all have the same structure.

for(i in 1:100) {
df <- data.frame(x=rnorm(100), z=runif(100))
write.csv(df, sprintf('file%d.csv', i))

Then we can write up a convenience function to load plyr, find all the csv's, and define a function that we will run based on the user-entered arguments.  Keep in mind you're going to want to pass the output of this function into a variable. <- function(dir, stringsAsFactors=F, keepMeta=F) {
files <- list.files(dir, pattern='\\.csv',
toExec <- "mdply(files, read.csv, stringsAsFactors=stringsAsFactors)"
if(!keepMeta) {
toExec <- paste0(toExec, "[, -c(1, 2)]")

You're probably wondering about the "keepMeta" argument. When you run mdply() without adding the column subset to the end of it you end up with two extra columns: one for the index number of the file it came from in list.files(), the second being the actual row number that record resided in within that file. I find that info to be unnecessary most of the time, hence the default of "keepMeta=F" re-writing the function so that it excludes the offending columns.

As a side note about this, I ran my function against a for loop to see how well it performed.  I believe the results are quite clear and provide yet another piece of evidence as to why you should vectorize your code whenever possible.  Additionally, the *ply functions in plyr support multi-core execution which could cause the function to execute even quicker.  This is an area I'll have to research perhaps for a "Part 2".  The for loop used is after the image.  In both cases, they were simply dropped into system.time() and the "User" time recorded.  It was done for 10, 100, 500, 1000, 2000, ..., 10000 files.


for (i in list.files('.', pattern='\\.csv', {
myData <- rbind(try(myData, TRUE), read.csv(i, stringsAsFactors=FALSE))

One additional thing I learned is that when testing for loops you really should write a script to automate it.  It took more time for all of the for loops to finish than it did to come up with the idea for this post, write the function, test the function, get badgered my co-blogger for taking forever to get this post up, and write all the portions of the blogpost not relating to testing the for loop.

Also, I apologize for the weird formatting.  I was hoping to figure out how to get WordPress to respect my code indentation, but it doesn't seem to agree with me on that.  Hopefully you can forgive this fact as a first post while we try to figure that bit out.

Leave a Comment

Filed under Functional Programming, R