This workshop covers the mechanics, rather than the methods, for performing a qualitative data analysis. The examples will use the RQDA package running under RStudio.
RQDA is available on both PRI windows application servers: Kiev and Moscow. It is also available on the PRI linux network. Note that you must ssh into finland to run it as it does NOT run through the webserver version.
If you want to load the software on your own machine, you will will not only be installing the RQDA package, but also GTK and RSQLite. Information on the procedure can be found on the RQDA website http://rqda.r-forge.r-project.org/.
Additionally, you will want to install the FireFox add-on: SQLite Manager.
Once your packages are loaded you will want to create a folder to store your text files. (In RStudio, you can create an 'RStudio Project' in this folder.) In addition to the text files, you can store any R programs and you will create an 'RDQA Project' coding database there too.
Now load the RQDA package by checking it in the Packages list in RStudio. The first time you load the library, RQDA should start on its own, but if you close it during the session you can restart it again from the command line:
RQDA()
The RQDA panel opens a separate window, but keep your normal RStudio console open as you will need both windows.
The main panel includes….
For a new project, click on the “New Project” button and supply a name and associate the name to the folder on your system that you created above. This will create a file on your system with the extension '.rqda'. Once the project has been created you can click the “Open Project” button.
You can backup your entire database easily. Choose a standard naming convention so that the backups are easy to identify on your file system.
You can click on the “Settings” button to set some defaults for your project. This is especially useful if several people are working on the same project because you can identify the person coding.
n.b. GTK display will likely throw errors. In most cases you can simply ignore the errors. On Windows, your theme could obfuscate the buttons. If so, change the theme.
There are 9 basic Panels
Each panel has a set of tabs at the top for the most common actions relevant to that panel.
Each panel also has a set of actions that you can access with a right-mouse click in the panel area. Some of the actions are exactly the same as the top-tabs, but there are additional actions that are very important here, so be sure to examine them.
You can include some project-level comments/documentation by clicking on the “Project Memo” button. Memos can be attached to other levels such as the Codes, Files, Cases.
Files are the interviews for each Case in your study. As of RQDA version 2.1 only plain text files are supported. Word documents or RTF files will need to be converted.
If you change the original text file in any way, you will need to re-copy the modified file again.
You have several options for inserting your text data into RQDA…
The content of your text files will be copied into a database when you import them into RQDA. To import a file into your project, click the “Files” button and then the “Import” button, then select the text file.
If you want to import many files at once, you can use the write.Files command. This is usually not an issue when you are coding individual transcripts, but it is a very useful feature when you want to extract open-ended questions from of a larger survey questionnaire for coding. For example, I have a large survey of multiple choice questions, but question 64 is an open-ended question “What do you do with your friends when you are not in school?” I can pull those responses into RQDA for coding. To begin, export the subject ID and open-ended question variables into a csv file from your SAS,Stata, SPSS, etc dataset and load the csv into an R data frame.
# load csv containing case ID and open-ended question into an R data frame
# named q64
q64 <- read.csv(file = "q64.csv")
str(q64)
## 'data.frame': 470 obs. of 2 variables:
## $ ik_64_open: Factor w/ 437 levels "a lot, sometimes they pick on miranda",..: 152 131 145 320 29 352 228 207 51 306 ...
## $ kidID : int 9 10 11 14 15 17 22 23 24 25 ...
head(q64)
## ik_64_open kidID
## 1 play basketball, watch tv 9
## 2 play around, hit each other, make each other laugh 10
## 3 play basketball, play video games 11
## 4 talk, watch tv 14
## 5 don't know 15
## 6 try to get staff to let them ride bikes and watch tv 17
# Insert file and contents into db Startup RQDA
library(RQDA)
RDQA()
# Open the project from the command line rather than the GUI
openProject("~/Documents/RStuff/QDAtm/QDAtm/QDAtm/q64.rqda", updateGUI = TRUE)
# Pull the text of the question contained in the column 'ik_64_open' from
# the q64 data frame
q <- as.list(as.character(q64$ik_64_open))
# Assign the subject ID as the name of each File
names(q) <- q64$kidID
# Here are the first 3 lines
q[1:3]
# Use the write.FileList command to copy this info into the RQDA database
write.FileList(q)
# Close the Project from the command line
closeProject()
The last way to create Case Files is to enter your text data directly into RQDA. Click on the “Files” button and then the “New” button. Enter a name for your Case and a text editor window will open for you to enter data. Your data will not appear as separate files and you will need to open RQDA to edit the data. However, we will see later that you can easily export the text from the RQDA database. Using this method you will not need to worry about file conversion. Note that to edit an existing file, right-click in the file listing area of the Files panel to locate “Edit File” menu option.
Click the Codes panel button and use the +Add button to enter your codes. Press “OK” You can add a Memo for each code to describe its usage.
You may have different file categories, for example an interview may be conducted as a face-to-face or telephone interview. You can associate a “File Category” to each file in order to pull files of a specific category.
If you have multiple interviews for one person in separate files, you can create a “Case” id to associate the files with a particular “Case” or unit of analysis.
Codes can be aggregated into Code Categories at any time during your coding. A code can be in more than one category.
You can add attributes like gender, age group, school, etc to a File or a Case. These are a bit different than codes or categories in that they are a variable attached to the case with a single value. Advantages to using them are that you cannot possibly put someone in multiple categories that should be mutally exclusive. Disadvantages are that attributes are not as easy to use through the GUI. When you are adding attributes ask yourself whether you really need this information in order to do the coding rather than the analysis. If you don't need it to code, you can always store the information in a standard data frame and merge your coding later.
You can view the results of your coding in a variety of ways. Retrieval is done in the Codes panel. However, overall options for retrieving are determined in the Settings panel. This genenal 'Type of retrieval' setting determines whether or not the selected values in the Cases and File Categories panels act as filters when you retrieve your codings.
From the GUI: In the Codes panel you can right-mouse on anywhere in the panel and select “Export Codings as HTML” You will be asked to enter a filename and you can select one or more codes to be displayed. A nicely formatted report will be created that you can open in your web browser.
You can also select a file in the Files panel to export as html with the codings View report
From the command line: You can request information from the command line in RStudio when the project is Open.
summaryCodings()
This will produce summary counts of the number of times codes were used, number of files using each code and the length of the coded text segments. You can store these summary statistics in an R data frame for subsequent analysis.
----------------
Number of codings for each code.
school talk teamSports
3 4 1
----------------
Average number of words assciated with each code.
school talk teamSports
22.66667 80.25000 10.00000
----------------
Number of files associated with each code.
school talk teamSports
3 4 1
You can create plots to describe your codings from the Code Categories panel. Select one or more Code Categories and right-mouse to select “Plot Selected Code Categories”. A new window will open with a sociogram of your codings. This is an active diagram and there are many manipulations you can make. You can export the final diagram to postscript.
You may never actually need to look at the Database that stores you coding, but it is very easy to do and it will help you understand how to use the software. The most common reasons to look at the database include:
# You will need to open project first and get codes and file names
myCodes <- RQDAQuery("select source.name as filename, source.id as fid, C.cid as cid, C.code as code\nfrom source inner join\n(select freecode.name as code, coding.cid, coding.fid from coding inner join freecode\nwhere freecode.id = coding.cid and coding.status = 1) C\nwhere source.id = C.fid and source.status = 1")
# Then need to reshape into file id (or subject)level
myPersonLevelFile <- reshape(myCodes, v.names = "code", timevar = "cid", idvar = "fid",
direction = "wide")
# Grab category info
myCodeCats <- RQDAQuery("select treecode.catid, codecat.name as codecat, treecode.cid\nfrom treecode left join codecat\nwhere treecode.catid = codecat.catid\norder by treecode.catid, treecode.cid")
If you entered your interviews directly into RQDA, you will want to have copies of the text in plain text files that are separate from the database.
[1] There is a terrific series of videos on using RQDA on YouTube http://www.youtube.com/playlist?list=PL66CB2FF65368715C. I highly recommend them.
[2] The official citation for the package follows: HUANG, Ronggui. (2011). RQDA: R-based Qualitative Data Analysis. R package version 0.2-2. URL http://rqda.r-forge.r-project.org/. Be sure to give the author credit in your publications.
# Create Corpus from a directory of text files
library(tm)
source <- DirSource("~/RStuff/qda_tm/QDAtm/QDAtm/responses/") #input path for documents
YourCorpus <- Corpus(source, readerControl = list(reader = readPlain)) #load in documents
tdm <- TermDocumentMatrix(YourCorpus, control = list(removePunctuation = TRUE,
removeNumbers = TRUE, stopwords = TRUE, wordLengths = c(4, 100)))
plot(tdm, terms = findFreqTerms(tdm, lowfreq = 16)[1:25], corThreshold = 0.25)