MATERIAL Y MÉTODO
5.1 VARIABLES CLÍNICAS Y MORFOLÓGICAS.
After completing all the interview transcripts, there was far too much data to analyse in single passes, with open coding being all but impossible. To manage this, I summarised the interviews into paragraphs, extracting just the information about sounds and sound- scapes into the activity-based categories established in Wave 1. This rst-degree level of integration gave me a much more manageable document to work with when, for instance, trying to nd out all the times people discussed a single topic or concept.
Coding-wise, I needed a new tool given NVivo's issues. As I am an experienced web developer, I elected to move to a custom, HTML-based system, migrating my interview summaries and the twenty interview transcripts to two HTML documents. I recreated the core feature of NVivo coding in a simple, stripped back way, that allowed faster access using more familiar tools, without platform or licensing restrictions. I plan to release this code at a later date for others to use. I will refer to the tool as `SQDA' `Simple Qualitative Data Analysis'.
Figure 4.9 on the next page shows the working environment. Sub-gure `a' demonstrates an interview summary in which some elements have been colour-coded to show their position on an arbitrary dislike strongly like strongly scale that is not in the nal analysis. The `People' button in the top left corner is a shortcut to the appropriate interview all twenty are listed in order for easy scanning. The `vol', `opinion', `place' and `tags' dropdowns in the top navigation are the ltering functions for data traversal. These are shown unpacked in Sub-gure `b', with the data currently ltered to only show things tagged `control' (visible in the background).
Technical Description
HTML5 contains `data attributes' which can then be manipulated using various JavaScript (jQuery3 specically) plugins. Using Sublime Text 24 (an advanced text editor) for data
entry, Google Chrome with its very fast V8 JavaScript engine, and a browser auto-reload plugin5 to update the document on save, I made a exible system with a very fast
response time. Also, using Sublime Text 2 allowed me to use other `tools of the trade' like regular expressions (highly advanced searching), advanced editing and highlighting, and the exibility to write my own extra code as needed. Twitter Bootstrap6 was used for a
basic theme, and jQuery Filterify7 for tag ltering. I used re.app8 to provide basic Ruby
3http://jquery.com/ 4http://www.sublimetext.com/2 5http://livereload.com/ 6http://getbootstrap.com/ 7http://luis-almeida.github.io/filtrify/ 8http://fireapp.kkbox.com/
(a) SQDA: view of an interview summary.
(b) SQDA: showing various lter options. `Control' is currently selected. Figure 4.9 SQDA: summary view
(a) SQDA: Interview markup denoting quotes which have been used with a used class assignment
(b) Summary markup showing data-* attributes Figure 4.10 SQDA: markup examples
templating and a local, portable webserver, allowing me to work on dierent computers, and synchronise the whole application with Dropbox.
In Practice
Figure 4.10 shows examples of the markup. In the top example, the interviews use simple .question and .answer classes to semantically markup the data. .used classes are added to keep track of interview sections that have already been copied into the thesis. The bottom example shows how the data-* attributes are used in practice. These can
Figure 4.11 SQDA: SASS markup showing ease of adaptation
be added or removed at will, and are comma-separated. The lter plugin then accesses these in order to hide things which do not have that attribute. Some jQuery helpers add line numbers and hide subheadings in which everything is hidden.
Various colour-coding was used on an ad-hoc basis. This was done using SASS9, a CSS pre-
processor that allows for clean, semantic code, as seen in Figure 4.11. The top li[data-*] elements denote colour coding for an arbitrary opinion axis. The bottom .used codes are increasingly opaque shades of blue (the last rgba value is opacity) to denote used interview sections the output of this is shown in Figure 4.12 on the following page.
Both interview summaries and the overall transcripts were parsed using regular expres- sions (regex) when the need arose. Figure 4.13 on the next page shows an example of a regex search for the string \s(bus(es)?|cars?|traffic|road)\s for example. This will return any string that matches, in this order:
1. A space
2. Any of the following:
Figure 4.12 SQDA: view of an interview transcript. Quotes which have been used are highlighted.
(a) `bus' or `buses' (b) `car' or `cars'
(c) `trac' (d) `road' 3. Another space
Bookending the search string with spaces like this prevents the search from returning unwanted, common hits like `card' or `busy' for example. This example could be im- proved: for instance replacing the last \s with (\s\.<)? would also return hits ending with a fullstop, or closing paragraph tag. In Figure 4.13 on the preceding page, the white border around ` bus ' indicates the next `nd' keystroke will cycle to that word making searching very quick.
While this solution wouldn't be suitable for everyone, the speed, simplicity (for me), and customisation paid dividends for fast and eective editing, coding, and theorising. As a platform, web browsers and tools are highly advanced. For example, I was able to add in extra features such as paragraph numbering and change the layout at will as my needs changed. Chrome and its JavaScript engine are extremely fast and could search the entirety of my interview transcripts instantly a vast improvement on NVivo's sometimes ve-second response time for every search or view change.
The ideal end-product for this would be if Chrome allowed local le saving. This would allow use of the browser as both limited editor and browser, whereas at the moment making substantive changes to the document requires the use of a separate program. This feature is not currently implemented, however.
Taking us back to the research at hand, this stage of data processing resulted in setting the stage for an analysis breakthrough. It resulted in creating access to code and category labelling via fast data traversal that was impossible or impractical using NVivo or Word. The process of creating data manipulation tools also resulted in the summary process, which in broad terms outlined listener proles.