Collecting experiential data

Last week I spent a little while at Bodiam Castle, collecting some pre-pilot base-line data on the experience there. This is a continuation of the Ghosts in the Garden research, testing some alternative questions and a different approach. At the Holborne Museum, I used paper surveys. This time, I tried a face-to-face approach. I had been planning on doing it all on paper, but as the date approached, and the weather looked wet, I decided to try a more technological approach. Some online research led me to QuickTap Survey. This is an on-line service that makes it easy to create a survey on their website, then download it to a mobile device (I used my first generation iPad), when it appears almost as a “kiosk”, with a screen for each question and very easy to use touch controls or on-screen keypads for responses. There is also an option to put the questions on one page, like a paper survey, but I didn’t try that out.

It turned out to be a great tool in the field, really responsive and quick to use. Each sample took less than two minutes to interview (I asked twenty questions). There was a slider for the Likert scale questions, that some (most) visitors were comfortable using themselves. This has great potential, because it allowed continuous responses. I used seven categories (0-6), but people were chosing to push it right to the top end of six, or “only just” into the six region. Given that the slider stays the same length no matter how many categories you use, you could easily create a Likert scale with 100 responses, to get something that feels like, and may actually be, a statistically continuous integer scale.

There are some working traps to be aware of: I created a question requiring a yes/no answer, but the only options for answers were true/false. But it allowed me to create multiple choice questions with either “select one only” or “select all that apply” and along with the Likert scales and a couple of numeric questions, I was good to go. I was at Bodiam for two or three hours on a damp day when it wasn’t too busy and I managed to ask almost everybody leaving the Castle to participate. None refused, so I only missed the occasional group who left while I was already engaged with a visitor.

The app saves all the responses on the device (which is good because there is no mobile signal at Bodiam) and you can upload them when connected to wifi. I did have a problem here at first, because it turns out there’s a (known) bug which means you have to tell the device that it lives in Canada before the upload works. It was frustrating at first, but the help team responded quickly to my email with a fix (change the iPad’s region settings for the upload).

You can only view that data once you’ve uploaded it. But once its there, you can see it on-line, look at some pretty but not wildly useful histograms and pie charts and, crucially, download the data to your number-crunching computer. The download options are excel or CSV. It looks like the most useful one if you are going to do any real work with the data is “raw CSV” which is mostly numerical. The others all include the actual category words “disappointing, very enjoyable” in the data, which isn’t going to be useful in R. The raw CSV file isn’t perfect though. The True/false data comes as 1 or 2 rather than the 0 or 1 which you might expect (though having typed that, I recall there may be a good statistical reason for that which may have been mentioned in my Coursera course). And he “all that apply” multichoice data comes as a single field with comma separated numbers relating to the order of the categories. An “enhanced CSV” file splits out those categories into separate columns but, frustratingly, doesn’t populate those columns with numeric values but instead repeats the category name. So it seems I’ll have to do a bit of fiddling before I can load the data into R and have a play with my newly acquired statistics skills.

All in all though QuickTap Survey seems a useful bundle of service and App. Its pretty expensive though. I used a Free level, which allows me just one survey with a maximum of 50 questions and 50 respondents. The next level up (which allows for up to ten surveys, 100 questions per survey and 1500 reponses per survey) costs $19 (CAD) per month, and additional devices (if you want more people collecting data) cost at least $9CAD per month each.

It may be that when I need to break the 50 response barrier, I can organise my work to get it all done in just one month, and there’s a free trial of any of the paid levels of service too, but I which there was an academic level for us poor students, like the one Prezi offers.

Now about those “newly acquired statistics skills”. I’ve got a mid-term exam due tomorrow, and this week’s coursework needs to be done by Sunday, So I”d better sign off.