Cross-eyed PhD: Tip 5 Scale Creation

 

“Piled Higher and Deeper” by Jorge Cham, www.phdcomics.com

 

The Significance of DATA in research…

 

In your research, you will be presenting empirical evidence substantiating your analysis. To do that, you need to get (valid, reliable) data. Do note that if it cannot be expressed in figures, it is not science – it is an opinion.

In a social sciences research like mine, you will be creating a scale to measure the perceptions of your target audience. With the scale you will be grading the levels of perception or ‘getting a scaled data’ for developing your model.

There are a number of books on research that capture the principles of creating a scale for capturing research data. I recommend that you get the basics thru one of these books. In particular, I recommend the following:

  • Handbook of Marketing Scales … Bearden, Netmeyer & Haws
  • Essentials of Business Research …Jonathan Wilson
  • Research Methodology…Chawla & Sondhi

As you look at the diversity of scales available in the prior learning, some common themes do emerge:

  • A number of items will be proposed for measuring a particular attribute – or a factor in that model. However, you need to be careful in selecting too many items as it will distort the reliability of the scale. Modern social science scales are crisp and light.  I started with 140 item instrument and I ended with a pathetic statistical reliability! However, the initial analysis recommended which items are redundant and eliminating them improved the overall Cronbach alpha to a statistically acceptable value – and the instrument slimmed to barely 30 items.
  • Likert scales are popular in social sciences. Of course, there are researchers who question their validity as an interval scale. Their use is with an underlying assumption of their capability to granulate the perceptions to make them much more than a simple ranking. Again – the literature seems to be divided on a 7 part scale or a 5-part scale. I experimented with the 7 part scale and found it cumbersome for the users and I finally used a 5 part scale.
  • Do note that at practically all stages of presenting your results, your academic audience will be checking if the scales are reliable and the data acquired relevant and valid...and the analytical process robust. It may appear simple – but be prepared to spend appropriate time here and make sure that the research process is meeting the academic rigor expected of an acceptable work.

A key trick I would like to share as a tip is my use of OMR (optical mark reader) form – and a sample is included here.

Sample OMR Form

 

The OMR form’s advantages are:

  • it delivers a neat layout and a visually appealing instrument as compared to some of the research instruments I came across.
  • It is easy to fill for a respondent (though care has to be taken to guide on the filling of the OMR dot)
  • …and its biggest advantage is that a properly designed OMR form is machine readable. That means, I could just fill my 4 page 500 respondent questionnaire into the document scanner and it would scan and extract my data file in less than an hour!

How did I do  this? I started by designing my OMR forms using Corel Draw and OMR fonts that I downloaded from the net. Then I used Gravic’s Remark OMR software.and scanned the form and created a software template that identified the number of rows and columns on the page and the significance of each dot that the software would read on the data matrix. Once I had templatized the soft questionnaire, the process was simple. I conducted my field survey and got all the filled forms. The forms are duly numbered and marked for each city/ location/ agent to enable an audit trail. These forms are fed into the document scanner( I used HP Scanjet 7000 with scan speeds of 60 pages per minute)!

The document scanner creates the scanned file and you can then review the data vs the scanned image for the typical errors – and correct the obvious ones to create the final data file!  The software will also guide you by identifying blanks, duplicates and several common errors. Typically, this part of the data extraction and audit would have taken a few weeks in a manual process. Further, the manual process is prone to errors and can be difficult to audit separately.

For more details – please see this video and feel free to contact me if there are any queries:

I owe this trick to my stint as the product manager for HP Scanners – and my domain understanding of the document management applications! Strongly recommend it to the newer generation of researchers – you can get vendors in most cities and you can really cut out the manual drudgery in this part of the process.

The final data file will be used in the analysis stage of your research.

And, once again – here is one more quote to inspire you on your research path: