Within mixed methods research, what you are doing is known as an explanatory sequential design, and I recommend that you read the chapter on these designs in Creswell and Plano-Clark's (2018) textbook on mixed methods..

I agree with your strategy of choosing "data rick" cases, but that will require insight into your results. For example, if your original hypotheses are largely confirmed, then you can draw on illustrative quotations to show how those hypotheses worked in practice. Alternatively, if you want to examine predictions that did not hold up, then you might want to choose outliers from the quantitative analyses.

Thank you, Dr. David L Morgan, for your reply

Yes, I am using an explanatory sequential mixed-methods approach, and I am using this book as a guide.

"Alternatively, if you want to examine predictions that did not hold up, then you might want to choose outliers from the quantitative analyses." Do you mean "Deviant Case Sampling" is the most suitable one?

I would like to mention that 10 hypotheses out of 28 were not significant. This led me to think about focusing on it

If you could mention any source that explains the process in detail, I would appreciate it.

Thank you again

Yes, this sounds like it is the same as "deviant case sampling," in the sense that both of them assume that there was a dominant pattern which had its effectiveness reduced by cases that did not fit that pattern. And by outliers, I mean cases that you could detect by applying a scatter plot to a correlation and regression line.

But, the data might be entirely random with regard to any given hypothesis, in which case there would not be any particular set of "rich" data points.

Thank you Dr. David L Morgan, for your clear explanation. It is really helpful.

The goal is to find participants who score:

LOW score DV-HIGH score IV and

HIGH score DV-LOW score IV

The cleaned dataset is normally distributed.

I used the following steps to identify the potential deviant cases based on the DV-IV difference that showed a non-significant effect:

1. Calculate the DV-IV Difference: Subtract each participant's dependent variable mean score from their dependent variable mean score.

2. Calculate the mean DV-IV difference: Find the average of all the individual DV-IV differences. It gives a central point of reference.

3. Calculate the Standard Deviation: Determine how much the individual DV-IV differences vary around the mean. A larger standard deviation indicates more variability.

4. Determine a "Typical Range": Use the mean and standard deviation to define a range of "typical" DV-IV differences. A common approach is to use two standard deviations from the mean:

Lower bound: Mean - (2 * Standard Deviation)

Upper bound: Mean + (2 * Standard Deviation)

I calculated for each DV-IV from the cleaned dataset. Then, it is combined with Maximum Variation Sampling

Are these steps correct? Should I apply a percentage (i.e., let's say 10% of the bottom cases instead of standard deviations)?

Thank you again for your comments.

This sounds like a complex procedure. Assuming that you have a non-significant correlation, I would generate a scatter plot with a regression line and then look for the two kinds of "deviant" cases that you described.

Thank you Dr. David L Morgan for your clear explanation. I will follow your suggestion. Thank you again.