Hi Tichaona,

A mixed-methods sequential explanatory design starts with a quantitative study phase then follows up with a qualitative study phase. The purpose is to use quantitative methods to test/prove some hypotheses in your theoretical framework, then utilise qualitative methods to dig deeper into the field in order to provide explanations to quantitative results that are ambiguous because of the limits of quantitative data.

With this logic, there should be at least two ways of design sampling strategies.

A) Qualitative sample chosen from quantitative sample.

Step 1: Design your quantitative sample. This sample is of course randomly selected so as to best explain your research hypotheses. How to sample depends on a lot of factors to which you must figure out such as what kind of population, budget, human resources, time, etc.

Step 2: Design your qualitative sample based on the sampling frame of your quantitative sample. It means that you choose some individuals/households/companies (or whichever unit analysis) from the sample frame to conduct qualitative exploration. These chosen must provide extra information that help clear up the ambiguity arising from your quantitative analysis.

The advantage of this sampling strategy is that you can combine the qualitative information of those newly interviewed individuals (e.g., bibliographies, stories, motivations, attitudes, experience with unemployment, etc.) to the quantitative data (e.g. socio-demographics from questionnaire such as age, jobs, etc.). For instance, you found out that there a handful of people who are unemployed but have a very high annual income. You qualitatively interview them and see that they have been received remittances from their family members just in recent years, which make their income high - but this kind of information was not covered in your questionnaire.

B) New qualitative sample with a focus on theoretical model's explanation

Step 1 is the same as the A strategy's. But Step 2 you must choose a new qualitative sample different from the quantitative sample. This qualitative sampling is based on the demand from your quantitative explanation. With quantitative data, you are trying to fill up a theoretical model but there are some information is missing. You outline this gap and try to use qualitative sampling to fix it.

For instance, you surveyed farmers' households in a value change analysis of rice production. Quantitative data show that the rice production has been cut short recent years by most households but you find no evidence why from quantitative data. You then need to sample different stakeholders in a full circle of rice production such as retailers who provide agricultural inputs, merchants who connect farmers with companies, companies, government officials who have executed recent policy changes, and so forth. This new qualitative sample will be important for you to construe what is going on in your quantitative data.

Hope it helps.

Best,

kien

What is the best way of sampling in an explanatory sequential mixed methods research study?

Sequential explanatory design in mixed method research is started off with quantitative then follow by qualitative research. During the 1st stage of quantitative research, you can apply random sampling in order to generalize to the entire population. Based on your quantitative findings, you can select certain samples / respondents conveniently / randomly (from the previous sample pool) to perform the 2nd stage of your qualitative research e.g. in-depth interview to solicit insight that you can't extract from the 1st stage quantitative research. For an example, you can refer to this article.

Article Relationship between Team Satisfaction and Project Performan...

Qualitative sampling is purposive in nature, so that the participants you select can serve a distinct "purpose." In the case of an explanatory sequential design, your purpose is to improve your understanding of the quantitative results, so you should design a sample that meets this goal. For example, you might select outliers from a regression analysis (or simple scatter plot). Or you might decide that you need to learn more about the differences between people who score either high or low on some key variable. The key point is that the results of your quantitative research should define the purposes that define your qualitative sample.

As Kien Trung Nguyen points out, you can do this either by recontacting appropriate members of your original quantitative sample, or by drawing a new sample that meets your purposive criteria.

Hi David, agreed the 2nd stage i.e. qualitative sampling should be "purposive" - the word that I was struggling to look for in writing my previous post. Thanks.

Thanks very much for the academically nourishing responses. I am sure that I am not the only one benefiting from these scholarly rich contributions.

Hello Tichaona

In the qualitative phase, you can apply a combination of convenience sampling and theoretical sampling, until you reach the saturation point (point where no new categories or themes emerge). In the quantitative phase, a sample that is representative of the population is desirable. Given the complex structure of current urban populations, simple random sampling is not possible, therefore, the sampling strategy can be multi-satage random sampling.

Regards.

In addition to the rich contributions from all above scholars, you need to make sure that each objective of your research is perfectly aligned with its corresponding phase in your seq. Explanatory Research Design phases.

Is it advisable to use stratified random sampling for explanatory sequential mixed-methods study ( quantitative part)?

The goal of the qualitative portion of this design is to maximize its explanatory power. That is why, as I explained earlier, the best choice is purposive sampling.

David L Morgan , is there a modified sequential explanatory design? Or is what I'm describing an embedded design? I want to do the qual interviews by selecting participants in the top quartile of my quant measurement. But, due to timeline issues (feasibility), I need to identify those for the qual after each 25% of the quant data is collected. So, after for example, 20 participants complete the quant portion (of a planned n=80), I identify the 5 with the highest score on the measure, then perform the qual piece with them. This would occur after each 20 participants. So not a true sequential explanatory as not all of the quant data collection would be complete. Thank you for any help!

Embedded designs are usually contained entirely within quantitative projects, and most of the available examples actually consist of explanatory sequential designs (i.e., the qual examines the QUANT results).

Since your design does not have a common name, I would look at Teddlie's 2007 article on sampling in JMMR, 2007.

In an explanatory sequence mixed method research design, quantitative research will be conducted at the initial stage, and then gradually moves to qualitative phase. Therefore, sample methods for quantitative research could be any kind of probability sampling (random sampling etc) whereas qualitative phase, you can use non-probability sampling such as purposive sampling or convenience sampling.

Thank you ..all responses throw light to various aspects of MM sampling. Is there a specific nomenclature system of sampling in the mixed method?

Roshan Panditharathna

In a QUANT-qual explanatory sequential design, we always choose participants for the qual component from those who intially participated in the quant phase. Since quantitive and qualitative samples include the same participants, this sampling technique in MMR is called identical sampling. In instances where different quantitative and qualitative samples are drawn from the same population, we consider this parrallel sampling. However parallel sampling best suits convergent/parallel designs

I too am struggling with choosing the design that best fits my research questions. Both convergent and explanatory sequential have elements that match my intention, so I can't work out which suits best. I'm working from constructionist approach, and wanting to use surveys and interviews to elicit rich data. Can anyone advise?

Explanatory sequential designs emphasize quantitative methods (QUAN --> qual). Because the quantitative results drive the goals and design of the qualitative follow-up, that study seldom uses a constructivist approach.

Thanks David...that helps clarify. My other dilemma is constructivist vs constructionist...seems Guterman (2006) provided these two perspectives:

Although both constructivism and social constructionism endorse a subjective view of knowledge, the former emphasizes individuals’ biological and cognitive processes, whereas the latter places knowledge in the domain of social interchange.

Constructivism seems the popularly used term, however, I think the debate about which to use has gone on for years, & will probably continue.

I think the differences depend more on an author's preferences rather than any core underlying differences.

Let's assume I use QUANT in my first phase . Can I use document review (financial statements ) in the second phase to understand the QUANT results?