One of the final steps in creating my survey instrument was to send it down to CSR for them to review and reformat it so their scanners can read it.  The main part of that process involved setting up their scanning program to read this instrument.  Not only did they have to indicate where to look for responses but also what the responses mean (i.e. a mark in this specific area is response number 3 to question 1).  This also involves telling the program how to record the responses (i.e. response number 3 to question 1 generates a value of “4″ for the “compuse” variable).  As can be surmised from the previous example, this setup process also includes naming and defining all of the variables that will eventually end up in the SPSS file.

Just as interesting and important as the automated processes are the manual processes that must be created, documented, and enacted.  Most of these are quality assurance or error checking processes.  For example, after a batch of surveys is scanned someone must manually review the places where the program is unsure (i.e. a large checkmark that spans multiple response boxes) or the response was too faint for the scanner to properly record (all “missing” values are checked to ensure they are actually missing and not a scanning error).  There are also a few points in the process where results are manually double-checked to provide quality assurance.

When the instruments are scanned, the data are inserted into a database.  Then the data have to be extracted from the database and inserted into an SPSS file.  Once the SPSS template is created (and checked and double-checked), inserting the data is fairly trivial.  It can get a bit tricky, however, if you’re merging in data from other sources.  In this instance, we’re merging the results from this survey with the results from these students’ BCSSE surveys but I’ll do that on the back end using SPSS instead of doing it on the front end with a database query; that will make it easier for me to merge these data into the institution-specific data files we return to participating institutions.  It’s also something I can do myself which gives me more control over and understanding of things (I don’t touch the database; that is all CSR).

There are a lot of small details not described in the above overview and I’m really enjoying learning about this entire process.  It’s nice that my survey is a relatively small one: ~1600 one-page instruments.  That allows me to be very hands-on which (a) ensures that I understand the whole process and (b) saves me money because I don’t have to pay someone else to do these things.

There are still some unanswered questions, mostly those surrounding what to return to participating institutions and when to do so.  I wish I had an answer to some of those questions but I don’t.  Part of this is caused by the fact that I spend almost all of my time working on NSSE or occasionally FSSE where data are collected and reports generated at pre-determined and coordinated times.  BCSSE, on the other hand, uses a rolling schedule where we generate reports and return data to institutions as we receive their data.  That might not sound like a big difference but it’s not just a different process but a different mindset, one I had not fully anticipated or appreciated.

Finally, it’s tremendously exciting to finally see data!  We’re going through several test runs to ensure everything is set up properly and I understand how everything works.  I’ve been able to glance at a handful of surveys during testing but it was finally real to me when I received the first (test) SPSS file with MY data from MY survey instrument.  It sounds silly to admit that a screenful of numbers is exciting and even exhilarating but it’s true.  I have quite a ways to go but through the haze I’ve glimpsed the light at the end of the tunnel reflecting off something shiny in the far, far distance.

In the meantime, I wait for surveys to be administered and returned.  I’ve mailed about 5,500 surveys to 11 institutions.  I was very fortunate in being able to get a fairly diverse group of institutions to participate.  However, I’m a bit nervous that a handful of institutions that have not yet committed to participating in NSSE in 2011 may decide (or be forced) to not participate.  That wouldn’t kill my study but it might require significant changes.

The final instrument was delivered to me not in the format I expected.  I thought it would be in a format that would allow me to easily perform a mail merge to insert the survey IDs for each survey.  Since this instrument is physically a separate sheet of paper, I have to be able to match this instrument to the main BCSSE survey instrument (because I’m using demographic information from BCSSE).  To do that matching, I am printing survey IDs on each of my surveys that correspond to the survey IDs already printed on the BCSSE surveys.  I am then physically inserting my survey into the BCSSE survey (which is a 8.5×17 sheet folded in half).  Having the same ID on both instruments will ensure that we can match the responses to these surveys even if the two instruments are separated after the students complete them.

The final instrument as I received it is exactly what is linked above.  It has a blank for the survey ID but how do I insert the ID?  More specifically, how do I easily insert the ID for hundreds of surveys?  Here is what I’m doing:

1. Create an Excel spreadsheet with the survey IDs I need to insert (each participating BCSSE institution is pre-assigned a range of survey IDs so I know them in advance).
2. Create a mail merge document in Word.  This document has only one thing in it: a mail merge field in the bottom right corner.  This document uses the previously-created spreadsheet as its data source and the field is placed on the document such that it lines up with the blank Survey ID space on the pdf of the survey instrument.
3. Complete the mail merge, generating a Word document that consists of a bunch of blank pages with survey IDs in the bottom right corner.
4. Print the mail merged document to pdf.
5. Open the new pdf and add the final survey instrument as the background of each page.

This process is not the most straight forward way to create a mail merged document but given my situation it’s not bad.  Although it’s a tiny bit convoluted it works reasonably well.  This process might not work if it had to be more precise (i.e. inserting new text into a block of existing text) but since I have a large target that is an empty box I don’t have to worry about precision.

(I was initially guided to this process – specifically the idea to use the instrument as a background image on another pdf – by this 2-year old thread on MacRumors.com.)

I’ve used this process to print the first set of 375 surveys to be mailed later today or tomorrow to an institution in New England.  It should work well for the other 3885 surveys I will be mailing to nine other institutions in the coming months.

Additionally, it occurred to me that I may end up with three groups of participating institutions:

• Web-only participant in NSSE
• Paper or mixed-mode participant in NSSE
• Non-participant in NSSE

The first group is my primary target group as they will allow me to easily calculate non-response as it relates to Internet access and use. But if I end up with institutions in the second or third group (and I won’t know for quite a while since NSSE registration isn’t open yet and won’t close until mid-late Fall) then I might be able to use them to do interesting “troubleshooting” since they will provide additional data. For example, institutions that participate in the paper mode of NSSE would offer me the chance to see what happens when respondents have the choice between paper and Web modes and how Internet access and use influences that choice.

The pilot did not go nearly as well as planned.  I was hoping to conduct 5 cognitive interviews and a pilot administration with about 50 students.  I feel far short of both of those goals.  I think that the biggest factor that contributed to this failure is the timing: I was so late in soliciting participants that we were nearly in finals and that is a bad time to get students to do anything.  I also limited my recruiting to one (large) residence hall in the belief that my requests were so simple and easy that I wouldn’t have to recruit in other places.  Finally, I think the advertisements were too plain and too prominently focused on the Internet, perhaps causing students to believe that I was recruiting only computer experts or geeks.

Despite the dismal participation in my interviews and pilot, I am pressing ahead.  First, my timeline does not have any slack time.  Second, the data I was able to collect was all very, very positive in terms of the construction of my survey instrument.  Finally, even if this stage in designing this new survey instrument did not go as well as hoped, all of the preceding stages went very well and were performed very thoroughly.  I take some reassurance from the idea that all of the stages that were within my control were done well and thoroughly (I hear my dad’s voice echoing in my head: “Control the controllables”), in part because there are some stages that I can not control.  I will, however, conduct more cognitive interviews this summer even as my instrument is in the field; it would be too late to make changes if I were to discover any problems but (a) I do not expect to discover any problems and (b) if there are remaining problems then I need to be aware of them.

So now my focus turns to recruiting institutions to participate.  Ideally, all participating institutions would meet several criteria:

1. Registered for the paper version of BCSSE
2. Administering BCSSE in on-campus events (orientation sessions, FYE classes, etc.)
3. Intending to participate in NSSE next spring
4. Possessing a diverse student body

The first criteria is non-negotiable as it is critical to my research design.  More than any other criteria, this one immediately and dramatically narrows the pool of potential participating institutions.  The second criteria is desirable as those institutions meeting it should be those institutions that have the highest response rates.  The third criteria is also non-negotiable as it is critical to my research design.  The fourth criteria is important because if I have a homogeneous sample – especially an affluent one – then I may not have enough variation to perform some of the statistical tests I would like to perform (in statistical terms I would not have enough “power” for operations such as logistic regression).

Getting all of these criteria to align – including the unlisted logistical one of “we haven’t mailed their BCSSE surveys yet” – is challenging.  In fact, I worry about the fourth one considerably.  I can’t change the BCSSE registrants to make them meet my criteria but I am placing a small safeguard in my study regarding the fourth criteria: I will be recruiting a few institutions who have diverse student bodies even though they may not be participating in NSSE next spring.  I will not be able to analyze NSSE non-response data from those institutions but by collecting demographic and Internet access information I will at least be able to compare their incoming student bodies with those of the other institutions in my sample.  If there are no important differences then I know that I’ll have nothing to worry about with respect to diverse student populations.  If there are important differences then either I will be able to explore it in my study or note it for future studies.

Since I am pursuing an opt-in strategy in recruiting institutions (i.e. I am asking if they want to participate and they must say “yes”), there is still potential for this study to fall apart if institutions ignore my requests or say “no.”  I was hoping to use an opt-out strategy (i.e. telling institutions that they will be participating unless they say “no”) but some of my colleagues are uncomfortable with that strategy.  In any case, this recruitment is another critical link in the chain that is this study.  From my viewpoint, in the middle of this mess and uncertain how or if it will all turn out, it seems like yet another fragile link in a long chain of fragile links.

I also finished the draft advertisements for my pilot study.  I first plan to conduct five cognitive interviews with undergraduate students here at Indiana University.  These interviews are a way to understand how respondents understand (or misunderstand) the questions and responses on the survey instrument.  In these interviews, the participants are asked to (a) read aloud the questions and responses, (b) verbalize any questions or thoughts they have as they read the instrument, and (c) think aloud as they answer the questions.  Asking the participants to think aloud – the “cognitive” part of the cognitive interview – gives us an insight into how they understand and answer the questions.  The advertisements for the cognitive interviews are rather simple and I will be compensating participants $10.

After conducting the cognitive interviews and making any necessary changes to the survey instrument, I will then pilot the instrument by administering it to 60 Indiana University undergraduate students.  As you can see in the advertisement (here is a different view), I hope to print out large (11×17) flyers, place pockets on them, and insert survey instruments into the pockets.  I will only be advertising in this one building so it should be very easy for students to grab a survey, fill it out, and return it for their $3 compensation.

I’m sure that both IU Residential Programs and Services (RPS) and IRB will want changes made to these advertisements.  But I’ve got good drafts and I am hopeful that they will not be changed much!  RPS is getting first crack at them because I am not quite sure if they’ll let me put surveys in the flyers so if they don’t like that then I’ll have to (slightly) change how I distribute the instruments.  Once I get the thumbs up from RPS, I’m off to IRB!

To establish the validity and reliability of the survey before launching it full-scale, I plan to conduct cognitive interviews with five Indiana University undergraduate students before conducting a pilot administration to 60 students.  So in addition to working on IRB documents I am also waiting to here back from our residence life folks regarding how and whether I will be able to advertise my interviews and survey in IU residence halls.  I hope to be able to advertise only in one building.  I work in a large building that is half residence hall and half office space so if I can limit my advertising to this building it will make it very easy for students to participate since they will only have to walk a few hundred feet to either participate in an interview or drop off a survey.  I hope to make it even easier for survey participants by including the surveys with the flyers.

This feels like a really big step as this is the first time I’ve sent out all of this information to so many people.  I feel like I’ve made a commitment now that I have asked so many people to take time to review these materials whereas before this was just all supposition and hollow talk.  It could still fall apart if IRB doesn’t like things, I have difficulty recruiting participants (I am paying them to alleviate this possibility), the interviews or pilots indicate monstrous problems with the instrument, or if this all just takes too long.  But I think I have a fighting chance, especially with all of these awesome people supporting me and offering feedback and assistance.

I’m very disappointed to have to do this and I haven’t yet figured out how this might potentially impact my schedule this summer.  But I have a lot going on in my personal and professional lives so this is a smart move even though it stings my pride and disrupts my summer a little bit.