It is certainly possible to start with a hypothesis and then design a way to test that hypothesis via qualitative methods -- for example by predicting patterns that will or will not be present in the data.
In this case, however, it sounds like you have created the hypothesis while analyzing the data. I think there are problems with first developing a hypothesis from a data set and then testing it on the same data set.
In the Popperian epistemology, your research -- whether qualitative or not -- cannot prove any theory (however many black crows you see, you can never prove that all crows are black). All you can do is find a likely hypothesis and show that it is wrong (i.e. falsify it). Note that finding an unlikely hypothesis wrong is trivial (this is called strawman's fallacy).
This is a very brief summary of modus tollens hypothesis testing.
Now, coming to your question, imagine that possible qualitative findings about your phenomenon of interest are data points on a nominal scale: A, B, C, etc. If it is feasible to differentiate between them in your observations, you can postulate your null hypothesis as A (it must be a likely choice based on your prior knowledge of phenomenon, see above). Then, if your observations point to B, you can claim that you falsified your null, and that is your original finding. (In lieu of statistical significance test, you can claim that the likelihood of misclassifying A, B and C in your setting is less than 5%; this is almost nearly equivalent to the coveted p < 0.05.)
Totally support the comment of Dr. Morgan above. You cannot state your hypothesis based on your findings. If is called 'begging the question', i.e. assuming your findings are right and then using them to show that they are right.
Statistical hypotheses about qualitative data can be formulated based on the frequencies of the values/items/categories.
Typically, null hypotheses are of the form: "the frequency distributions for the observations within the groups are obtained by sampling a population with the same common frequency distribution", or, somewhat simpler, "the frequency distribution for the observations (of this one group) is obtained by sampling a populations with this (given) frequency distribution".
Testing frequency distributions is achieved via the sums of squared and normalized frequencies that are termed "Chi-squared" values (χ²). The distribution under the null hypothesis (H0) is consequently known as "Chi-squared distribution" and can be used to calculate a p-value as P(χ² > χ²observed | H0). Tis is called... yes: Chi-squared-test.
I second David that it makes little sense to first look at the data (event to first do the experiment) and then think about a hypothesis test. The p-values -although it can be calculated technically - has not the expected meaning. Firstly, it will not control a type-I error-rate, and, secondly, even if one hopes that it roughly does so, this control is of no use if type-II error-rate is unspecified (what, in turn, requires to specify a minimum relevant effect size and to plan the whole study accordingly).
Instead of testing it would be more informative and helfpful to formulate some panel of relevant as well as of irrelevant effects and simulate data under these hypotheses. Then you can compare your findings against the simulated data and see
a) if you can distinguish relevant from irrelevant effects and
b) whether or not your data looks (much) more like showing a relevant effect
Stating relevant effects needs subtatial knowledge, expertise, and brains. This is nothing where a statistician could help you.
Generation and of hypotheses through qualitative methods is difficult and testing of hypotheses is even more difficult. if you are at initial stage of your research i suggest you to get periodic guidance from seasoned researchers conducting qualitative research.
Qualitative research is exploratory in nature and goes into the realms of unknown. Hence hypothesis to conduct research is difficult. However you can develop hypothesis with your qualitative research which can be tested subsequently
It is counterproductive to do an hypothesis in qualitative research. Even though, in several approaches you can even start with sensitising concepts, in general, researchers are encouraged not to have preconceptions before data analysis. Therefore making hypothesis is a bad decision to make. Nevertheless you are able to do prepositions when you already have discovered your core categories, realising how these categories interplay.
It maybe possible under Grounded theory but research tradition in Qualitative approaches prohibits the usage of hypothesis. This is due to the fact that qualitative approaches does not perform seek causal relationships.
By this account we should not pose a research question either. The difference between qualitative and quantitative research is that the former deals with meanings, while the latter deals with models. There is nothing counterproductive in having a preconception about a meaning of the phenomenon you study.
Why not simply ask your participant: is this the meaning that you attach to this phenomenon -- and then carefully listen whether the participant generally agrees or generally disagrees -- and then explore all other meanings he or she may develop.
Thus your hypothesis not only focuses your own inquiry, but also stimulates the participant to explore the subject further. On the other hand, when you come to your participant deliberately as a clean slate, you risk that he or she won't even bother to start explaining their meaning. Just a thought...
On the one hand, if you want to test 'a' hypothesis you will probably set your questions to offer evidence of several competing hypotheses as well as your own. Whatever your questions are, there are implicit assumptions or hypotheses. In other words whatever you might guess could be the outcome of a study your task is to remain aloof and try and tease out what's actually the case. On the other hand, you might well derive a hypothesis based on your qualitative research for further testing in the field by you or others. I have done this in a mixed-methods approach where subjects are asked a set of questions but the results are triangulated with other data in print, by other researchers or derived from other primary research with a different set of informants.
All social research deals with meaning - only the ways in which this is done differs between approaches. The difference between qualitative and quantitative research is that the former uses qualitative methods and the latter uses quantitative methods. These methods enable different kinds of research. It is quite funny to see how these different kinds of research are turned into ideologies that cast certain ways of doing research as the 'enlightened path', 'forbidding' things the other kind of research does. Intellectually, this is boring.
We always work with assumptions about our empirical object, some of which are theoretical. Grounded theory acknowledged this after a wrong start. In this sense, we all work with hypotheses - constantly. Making them explicit and being ready to overturn them any time is a good way of starting any research project, including a qualitative one. More specifically, qualitative research can refute any hypothesis of the kind "all swans are black" by providing a single observation of a white swan.
More interesting are hypotheses about causal mechanisms at work inour empirical field. I think one specific contribution of qualitative research is its potential to reconstruct causal mechanisms, which in most cases means that we need to inform our empirical research by some hypotheses about causal relationships and causal machanisms. These hypothesis do not take the form of hypotheses testet by quantitative research.
Thus, my answers are:
Is it possible to set hypotheses for qualitative data? Yes.
Is it necessary? To the extent to which hypotheses are necessary to explicate the prior assumptions we take into our research - yes.
Is it possible to test hypothesis with qualitative data? Only in specific cases (see above).
I agree that one would not begin with hypothesis if the goals of the research were exploration and discovery, but not all qualitative research has to pursue those goals.
As for using hypotheses, I mentioned earlier that one could begin by predicting patterns that would or would not be present in the data. For example, if are you indeed interested in meaning, then you might predict that men and women will assign different meanings to the same event. Or, if you are doing comparative case studies, you might predict that a given set of factors will have more influence in one set of sites than in other sites.
I also agree with Jochen that "We always work with assumptions about our empirical object, some of which are theoretical... Making them explicit and being ready to overturn them any time is a good way of starting any research project, including a qualitative one." No one enters the research setting as a blank slate (tabla rasa), so it is rather better to be reflexively aware of one's prior assumptions, and many of those assumptions can easily be stated in the "if... then... " format of hypotheses.
Yes, although the inclusion or exclusion of a hypothesis does not mean the research is good or, more importantly, helpful! I see you already have some outstanding responses above from Nelson, Glaser and Morgan!
If the qualitative study is exploratory in nature then we can go with research questions only. But on the bases of theoretical and conceptual framework of the problem we can assume certain tentative solutions and also can make the predictions. I think there is no hard and fast rule regarding making hypothesis in qualitative research. Rather it depends what is the nature of the study we undertake.
This is a interesting question. I would argue that in some cases - it is possible to and necessary to deploy a hypothesis for qualitative research. The hypothesis would be drawn from and /or informed existing knowledge in the area of study
Interesting question, and appreciate all the answers are empirical and logical. I think it is possible to test hypothesis through qualitative methods. Qualitative data may be meant to describe the nominal and ordinal data, where these are without any debate, testable with hypotheses. The concern here, is to formulate, state, test, refute or falsify a hypothesis with qualitative research method, exceptionally grounded theory is basically a hypothesis generating qualitative method.
Hence it is possible for all that. The lit. following may help:
Visually Hypothesising in Scientific Paper Writing: Confirming and Refuting Qualitative Research Hypotheses Using Diagrams