About Surveys

Recently I had the opportunity to travel to Japan and conduct several cross-cultural psychology experiments. For the majority of our experiments we used pencil and paper surveys. What a pain! We were collecting data from several colleges around the country, more specifically Hiroshima, Tokyo, Nagasaki, Aomori, Saporro, Kyoto, Sendai, and Osaka. Luckily we had about seven research assistants because it took forever to put all the information into a spreadsheet and then analyse it in SPSS. After putting all the information in the computer we also had to seperate different surveys from one another. It took about two months to complete. One of the other experiments, studying surnames and affluency, was completed using  an online survey tool. Although we had a lot of respondents in total our team probably only spent a couple hours having to analyze the results. It was so much easier to simply export the data into SPSS, rather than having to code, type, and then analyze. It saved us loads of time.

As the online survey software market grows, so do the features that are offered with various programs. Most have capabilities in other languages, depending on the language installed on the respondents computer. Many can even do multiple languages in the same survey. This is especially helpful when collecting data with a multilingual population. You can give the respondents the survey in the language that they would prefer. Thus, decreasing respondent comprehension errors.

Online surveys are no doubt the best way to do surveys. I was doing a survey for a project in my marketing class.  I proof read it and then I had some of my team mates proof read it.  It all looked good when we sent it out to our panel, but the next day I got an email from one of our respondents who Notified me that there were a couple of spelling errors (really embarrassing ones) in our survey.  Now in a paper survey there would have been no hope.  Reprinting and redistributing 50 or so surveys that had not yet been taken wouldn’t have been possible.  But because it was an online survey we were able to correct the errors in a less than a minute, and no one who hadn’t already seen the survey ever knew the difference. 

 Later that same year in my Economics class we were doing another survey to determine a demand curve.  About mid survey I was looking at the data, as it was being collected in real time, (which by its self is reason enough to use online surveys), when I noticed that one of our questions was not the correct question type to get us the data we needed.  The question was collecting what peoples preferences where but respondents were not ranking their preferences as we needed.  No problem I simply changed the question mid stream and we moved onward with survey, this time getting the data that we needed. 

This is the big benefit of online surveys.  You can see the results in real time and you can fix any errors in real time.  Online surveys are just so much more flexible and resilient than are paper surveys. 

Branch Logic works to navigate a survey respondent to a specific block or group of questions based on how they answered previous questions in the survey, or on certain data specific to that person (e.g., gender, age, ethnicity, department, class, major, etc.) For example: If a survey respondent indicates that he or she is a Psychology major, a block or group of questions may be created using Branch Logic will contain questions that are specific and applicable to Psych majors. Each person who indicates they are majoring in Psychology will navigated to this branch. Those who indicate a different major, will be taken to another branch created specifically for them.

Joy. Peace. Unadulterated Happiness, these are words that researchers often use to describe their mood when they are about to launch a project. Months of hard work and planning has come to a few simple days, and yet often one runs into bumps and problems. This happened to me the other week, when the night before the survey was to be launched, I remember I hadn't randomized any of the scenarios, participants were supposed to see. Stricken with panic, I quickly logged on to our online survey software, and luckily, found that the program could randomize it for me. No messing with latin squares or other annoying, vexing randomizing procedures. With a couple clicks of the mouse I was able to save myself from professional calamity.

The best part about my online software was I didn't even know which participants got which scenario until the results poured in. It made for a wonderful double blind experiment.  I'm now convinced that online survey software is the best way to randomize variables in a survey.

Max-Diff is a method of scaling in which respondents are asked to identify the most important attribute and least important attribute for a set of attributes.  It counts down significantly on the amount number of questions that need are asked in comparison to the Paired Comparisons technique.  It is based on a measure of customer choice and trade-off, instead of typical rating scale responses.  It can be used to generate importance or preference scores for multiple items such as brands, concepts, or attributes.

Basically, it works like this:

A respondent is shown a set of attributes;

      A  B  C  D

And are asked to identify the most important attribute and the least important attribute (They say A is most importand and D is least important).  From this one question we know five of the six paired comparisons:

         A>B, A>C, A>D, B>D, C>D.

The analysis of Max-Diff can be done using a number of different algorithms and from these we can estimate utility functions.  Probably the most commonly used algorithm to analyze Max-Diff is using a Hierarchal Bayesian procdure.  Hierarchal Bayes is beneficial because it allows for borrowing across the data.

 Qualtrics surveys is developing this functionality.

Multivariate data analysis is the analysis of multiple variables at the same time. This type of analysis is used to find how a set of variables explain one or more other variables. For example, sets of variables may explain one overall variable (brand loyalty) or may differentiate between key market segments. Similarly, a set of brand attributes may be used to map relationships to the key brands competing in the marketplace, thereby showing the strengths and weaknesses of each brand.

Some typical applications of multivariate data analysis are:

• Quality optimization (food, beverages, consumer products, insurance).
• Optimization of brand attributes.
• Multi-item Scale Development.
• Optimization of scale measures and methods.
• Classification of respondent and market segments.
• Development of new advertising and promotional materials.

While conducting survey research, most academic and private sector organizations will adhere to the code of ethics and practices established by the American Association of Public Opinion Research (AAPOR). This code calls for honesty, respect, and integrity in dealing with respondents, clients, and the public.

In this code of ethics, respondents should be given the content, sponsorship, and purpose of the survey so that they may make an informed judgment about whether they wish to participate. Any assurances, such as confidentiality or anonymity, must also be kept by the researcher.

Additionally, researchers are called to disclose fully to those who sponsor surveys the limitations and shortcomings of the survey and to avoid use of methods that deliberately introduce bias into the results. A survey report should include information on who sponsored it, who conducted it, exact wording and sequencing of questions, description of the population and how a sample was selected, sample sizes and sampling tolerance, and the method place and dates of data collection.

This information is seldom available in published research reports or media summaries, but should be obtainable with a phone call or letter to the sponsor of the survey. Today, online survey software has made this process of gathering the ethical information extremely easy. With a good survey software program, one can provide the relevant information and adhere to the code of ethics and practices with ease.

Qualtrics provides great data analysis.  Its survey software is constantly updating in real time, so results of a survey are constantly being reanalyzed and updated.  Some examples of the data analysis done by this survey software are: basic descriptive statistics, cross tabulations (chi-squared, etc.), graphic displays, and more.  I like it because it is much less work than calculating it out by hand, and it doesn't need to be exported to another program to do the initial data analysis, though it may be exported for more in depth analysis.

For most companies and businesses, obtaining feedback from customers and clients is essential.  It provides them with a glimpse into what they are doing right, what they are doing wrong, as well as ways in which operations and relationships could be improved.  But obtaining and organizing this feedback in a way it can easily be sorted and delivered to the right people within the company can often be challenging. 

Qualtrics, an online-survey company has the ability to tackle this challenge.  They have easy to create surveys and a lot of functionality.  You can make a survey that begins with asking the customer what topic best relates their subject of feedback.  And then you can provide them with space to jot down their feedback.  The neat thing about Qualtrics is that you can have the feedback sent to specific people based upon what topic was selected.  They call it the "trigger" option, because a selection triggers the email.  Lets say you make cellphones and you want feedback about the "cellphones breaking" sent to one group of your employees, and "troubleshooting" issues sent to another group.  When they select the "troubleshooting" topic, then that feedback will be automatically emailed to eveyone who has been pre-selected to be on that list.  It is really easy to set this all up and is great for organizing customer satisfaction feedback so that important feedback doesn't get lost.

Recently I was trying to figure out a way to do Least Square Regression without having to use matrices.  I looked for quite awhile online for some ideas on how I could do this.  Using Matrices to tackle regression problems become a problem themselves because the matrix computations required to determine regression take a ton of computer power and often bog down systems.  It is just too much for the computer to do quickly. So a way around this is by using summations.

With Least Squares Regression we are trying to determine our intercept and our coefficients of x.  You do have to be familiar with summation notation to understand this.  What we are basically doing in this approach is trying to minimize the difference between are dependent variable, y, and our function of x, f(x).  The summation notation for this is:

∑[y-f(x)]^2=minimum

Because our f(x) is essentially f(x)=y=a+bx, so then we can have a summation of:

∏=∑[y-(a+bx)]^2=minimum