Survey Tabulation

[Madapple]

I have recently been put in charge of developing a way to tabulate the results of an onlin epoll. All of the fields in the poll are free text entry -no check boxes, radio buttons, multiple choice type stuff, just short answers. How can I import the information from an excell document and have it tabulate the results for 106 fields?

[Lee Smith]

IMHO, you don't. Somebody needs to rethink this survey.

Good Luck

Lee

[Madapple]

sadly it's a bit late for that. This is an anual "best of the city" poll, it's already online and running so there is no chance of rethinking it!

is there no way to have filemaker automatically find and count duplicate entries?

If i have to set up 106 seperate calculation fields I will.

Or is there perhaps an "index" type of thing that will count the total number of times a word appears?

hmmmmmm and for that matter is there no way to import an excel document and hav eit use the first line of text as the field names - Actually creat a database with those fields...

[Ender]

Summarizing the data is not the problem. But having wide open text responses is. You'll get all sorts of variations of the same thing, and they won't summarize together.

Here's a recent thread about summarizing survey data:

http://www.fmforums.com/threads/showflat.php?Cat=0&Number=118258

[Lee Smith]

The problem is, you have the potential of an infinite number of responses to the 106 questions. And, the idea of a Keyword won't work, because they can be rendered useless or reversed in meaning with a modifier such as "Not".

You are probably going to have read each response, and classified the responses in order to meet your needs.

Sorry, I just can't see any way around your structure. If it were me, I'd pull the survey off line now, and replace it with one that has the structure you need. Take a look at Dragon Web Surveys By: Waves in Motion at:

http://www.fmfiles.com/dev13.html

HTH

Lee

[DanBrill]

I have formal training in and have done a fair amount of survey research. Open-ended survey questions certainly have their place. I just don't think their place was in the type of research you are doing.

Unfortunately, you'll need to go through each survey and look at each answer. A data coder will have to read the answer given and turn this into something more quantitative so that people who are giving essentially the same answer but in different words are coded has having given the same answer.

The main problem with this approach is known as "inter-coder reliability". different people reading the results should draw the same conclusions and code the data consistently. This is easier said than done. Before you start, you need to create a well-thought-out rubric of what the possible answers mean, and then train the data coders so that they understand and apply the rubric consistently.

Even if this all falls to one person, you need to check to make sure that your standards aren't moving over time as the fatigue of coding hundreds of these sets in. Anyone who has had to grade answers to essay questions knows about this. You want to apply the same standard to the first test as you do to the last so that a student'd grade isn't dependant on where their paper happened to end up in the pile. I'd hate for someone to get a lower grade on a question just because it was getting late and I was getting tired of reading the same crap 100 times.

Now since the type of opinion you are getting for a "Best in City" type of survey aren't likely to be all that nuanced, your job is simplified quite a bit, and there are a few things you might do from a database point of view to help speed up the coding. Suppose you have an open-ended question for best ribs. (Seems like a Memphis thing.) Call this BestRibs. Create another field called BestRibs2. Now search BestRibs for a restaurant you know will be in there somewhere based on your extensive knowledge of Memphis ribs. For any records that come up in the resulting found set, perform a Replace on BestRibs2, putting in the name of the restaurant you just searched for. Repeat this for another restaurant. Continue until some restaurant has a plurality of the votes. Anything that is left over can be marked as other.

Hope this helps,

Dan

[Madapple]

well luckily right now we only have about 105 responses to the survey, so it won't take very long to hand sort them. My bosses are wanting us to move to a strictly online poll and want me to develop a process to tally the votes for that time.

I found that if I simply open the excel document and do a sort for each column I can simply go thru and make sure all the "Rendezvous Ribs" are spelled correctly, and all the "Corky's" are entered as "Corky's" etc...

You are correct, there is indeed a category for "Best Ribs" although that category really only ever gets 4 major entries with any sort of consensus vote!

Sadly all 4 of those "Best Rib" establishments have nasty ribs, as any true Rib lover will attest to!

I wish there was a different way in which to administer the poll, but we can't have multiple choice type answers, just to ******* many rib restraunts around!

THANKS THOUGH!