If the pre-determined coding system is not related to the content of the survey, then it is arguably independent. Still, the general approach in convergence would be to analyze each of the datasets in terms of its own goals. Hence, you will have a stronger argument in favor of the separation of your two studies if you use an inductive coding approach that is based entirely on the qualitative data itself. If you do not do that, you risk the criticism that you "forced" the qualitative data into a scheme that was overly influenced by your prior hypotheses.

Deductive approaches to coding use a pre-determined coding system, while hybrid coding approaches begin with a pre-determined coding system and then expand and revise that system based on what is learned while applying it.

Either of these approaches might well be a poor fit to a pure convergent design, because they would violate the principle of collecting and analyzing the two data sets independently. Instead, I would recommend coding the qualitative data inductively and then comparing the themes you produce to the quantitative results.

You can find a discussion of this issue in my publication "Commentary—After Triangulation, What Next?" which is available here on ResearchGate/

Dear David L Morgan thank you for your response. what I am not sure I understood is when you say that it violates the independence of the data sets. I started both methods with the same hypotheses, so my themes are based on those hypotheses, not the survey or its results, which J haven’t analysed before the qualitative analysis. Therefore, is it still violating if I used a deductive approach and pre-determined coding (while of course staying open to new emergent codes)?

If the pre-determined coding system is not related to the content of the survey, then it is arguably independent. Still, the general approach in convergence would be to analyze each of the datasets in terms of its own goals. Hence, you will have a stronger argument in favor of the separation of your two studies if you use an inductive coding approach that is based entirely on the qualitative data itself. If you do not do that, you risk the criticism that you "forced" the qualitative data into a scheme that was overly influenced by your prior hypotheses.

I understand David L Morgan

Thank you so much.

If you have any time to read the following, I have described my work in more details so you could tell me if I did this right, or at least like you said in an "arguably independent" manner.

What I have done is this: I started with an extensive literature review, then I proposed my hypothesis which are also my research questions. From there, I designed my survey with close-ended questions (Likert Scale) about items related to the "themes" of my hypotheses and in parallel, I developed more general (not to lead the answers) open-ended questions for the semi-directed interviews. I analysed the qualitative data first to make sure the results from the survey did not influence my interpretation and then I ran parametric tests on my survey data. Finally, I compared the findings from both strands side-by-side. There were a lot a differences, which I discussed in my discussion and used to propose future research suggestions.

Ok you need to see the hypothesis.For example in case your hypothesis is that variable a is correlated to varaible b then this is a hypothesis which will be proved by a quantitaive technique.For this the questionnaire(structured multiple choice ) will be coded with numbers ,say Yes becomes 1 and No becomes 2.This is quantitaive coding.

Now come to the PI or qualitative or content analysis or text analysis of the open ended PI that you collected.Here the coding means naming the major variables for which you asked an open ended qs for example "what is the frequency of teh purchase" so here the frequency of purchase becomes a code and we analyse teh data of all samples who answered to this code (or open ended qs).