Dear Robert

I use NVivo software: I import the data, use emergent coding, and aim for a thematic analysis. Many universities have NVivo software available to use; if yours doesn't, the downside is that it is expensive. If you can get access to the software, there are good tutorials on YouTube, and useful books too - Pat Bazeley's is particularly helpful. Other qualitative analysis software packages I know of include Quirkos and MaxQDA. I haven't used either of these, and don't know how much they cost - though I think less than NVivo - but I have heard good reports of them from other researchers.

I think it depends on the kind of open-ended data you collected and the purposes it is going to serve. Helen has provided a good answer for questions that provide a fair amount of detail where you want to do relatively complete qualitative analysis. Alternatively, if you have brief data that you want to count and convert to quantitative variables, then it would be better to use content analysis as the basis for your coding.

All of the major qualitative analysis programs provide facilities for doing this kind of work, including NVivo, MAX QDA, ATLAS-ti, and Dedoose. I believe that SPSSS also has a special add on module that you can purchase if you are primarily doing content analysis. Note that most of these programs call this kind of analysis "mixed methods" since you are combining qualitative and quantitative data/

Previous commenters have focused on the software tool - I'd like to address the question about the approach taken to the analysis of open-ended responses in qualitative research.  The most common approach is Theme Coding, in which key themes are assigned (through review of each response) using a coding system.  The coding system may either rely on a Grounded Theory approach, in which the codes emerge from the data (as Helen Kara has described), or you make use of a pre-established set of codes that you have reason (theoretical, perhaps) to fit your data.  I would say that the former is much more common.  An offshoot of the  assignment of Themes is to treat these not just as a simple set, but to arrange them hierarchically, in which lower-level themes can be fit under higher-level ones, to produce a type of tree diagram to represent your findings.  This can be very interesting and useful -- a recent book by Miller et al. (2014) on Cognitive Interviewing Methodology gives good examples of each approach. 

Hi,

As one of the most popular software is NVivo 10  for Windows helps you to easily organize and analyze data in documents, and spreadsheets. 

NVivo 10 for Windows also can help to quickly arrive at answers, justify findings and make informed decisions. 

There is a tutorial and learning materials for you to understand how it work

Good luck

You can also use LIWC, which has a lot of research support and is widely used. A new version will be released this year.

http://www.liwc.net/

As others have said, it can depend much on the type of data you've collected and what you intend to do with it.  But my team typically takes a low tech approach and with a spreadsheet or SPSS, we create a new recoded variable or set of variables that lets us recode the qualitative responses into categories we can count quantitatively.

Hello Robert

Have you read this?

https://en.wikipedia.org/wiki/Computer-assisted_qualitative_data_analysis_software

Dear Jesper,

Thank you for the Wikipedia article!

It contains a very useful software comparison table.

Thankfully,

~Robert

How do you describe the approach when you use a qualitative coding method and then quantify the frequencies ?

Disclaimer: I'm the developer of the mentioned tool.

As already laid out by others, one the usual approaches is to map answers to one or more codes, which then allows a statistical interpretation of the responses. Machine learning can help in both generating a codebook (unsupervised methods) as well as in predicting the codes based on already annotated samples (supervised learning, svm, neural nets etc). For the latter we have developed a web tool (https://codit.co) which helps annotating answers efficiently by suggesting codes through a model trained in the background.

Gordon B Willis Hi Godon, thank you very much for your answer. I want to ask what to do if I don't have any "Grounded Theory" to code the open-ended responses? For example, in my survey, I asked construction professionals what barriers are to prevent the use of drones for construction safety inspection. Then, I got 300+ responses, and I just created the categories by finishing reading all the responses. In this case, how to validate the credibility of the categories?

Thank you again!

Yuting

I would say that you are in good shape with what you are doing: Grounded Theory involves 'grounding' the analysis within the data - in other words, you don't start with a theory, but derive whatever theory or organizational scheme directly from the data. So... not having a theory to start with is a GOOD thing, according to a Grounded Theory approach: You start with no preconceptions, and just rely on the data at hand. Does that make sense?

Gordon B Willis Thank you very much for your reply.

My paper partner proposed that maybe I need to ask other people to validate the categories, and I don't know whether this is necessary. The reason is that other people will never know the data as well as me, I mean, they will never check all the 300+ responses, then, why can I reply on them?

What do you think about it?

Thanks again.

Yuting

I am way to late to help Yuting Chen, but I would like to respond anyway, for other people with the same question. There is a lot that can be said for letting more than one person do the coding and comparing the coding frames that they come up with. This is usually done to assess and improve inter-rater or inter-coder reliability. However, not all researchers think that this is necessary, or even appropriate for qualitative research. In a recent publication (2020), O'Connor and Joffe describe the arguments for and against, as well as suggestions how to do this in practice. The article can be downloaded from Article Intercoder Reliability in Qualitative Research: Debates and ...

Gordon B Willis Charles Picavet , i have a questionnaire that consists of close ended and open-ended questions , initially i thought i could just take a quantitative approach but now i am thinking of taking both a quantitative and qualitative approach. I am a bit confused especially because this is my first time undertaking a research. It is a bit overwhelming for me. I would appreciate your expert advice.

Fredina Addo You might consider what mixed methods researchers call an explanatory sequential design (QUANT --> qual). In that case, you would first determine the results from your primary quantitative data, and use the open-ended questions to help understand the reasons behind those results.

Am really confused about the difference (s) between content analysis and thematic analysis as used in qualitative research. Can somebody help me?

Content analysis typically uses a deductive approach with a predetermined code book, and it emphasizes counting codes. Thematic analysis uses an inductive approach that develops the codes from the data, and emphasizes interprettation

Legborsi: You ask a very good question, about the difference between 'content analysis' and 'thematic analysis.' First, keep in mind that different researchers tend to use terms differently - or use different terms to mean the same thing. So, there may not be a substantial difference here (or there might be...), depending on who you ask.

But - I think a very good description of 'content analysis' is provided by https://www.publichealth.columbia.edu/research/population-health-methods/content-analysis. Based on this description, I see content analysis as more general, such that it includes thematic analysis. The authors break content analysis down into 'conceptual analysis' and 'relational analysis.' I consider conceptual analysis to be, in effect, thematic analysis (identifying themes, often through coding). Relational analysis is then the additional step involving establishing the relationships between concepts/themes. Take a look at the Columbia website, and it may be clearer than I have been here.

Grounded theory is always inductive. It uses as few preconceived ideas as possible.