Lecture note on verification and falsification in quantitative data analysis: for MBA students

Lecture note on verification and falsification in quantitative data analysis: for MBA students

 

In the context of analyzing inferential statistics findings, e.g. from a questionnaire survey on my Hong Kong friends' opinion on CEO branding", how to practice verification and falsification on quantitative data analysis, in the context of doing an MBA dissertation projects. Please highlight 4 ideas on each of these two concepts (i.e. verification and falsification) in your discussion.

In an MBA dissertation using questionnaire data, verification means strengthening confidence that your statistical findings are credible, while falsification means actively trying to disconfirm your hypothesis rather than only looking for support. For your CEO-branding survey, the best practice is to treat every inferential result as a claim that must survive checks for consistency, sampling quality, and alternative explanations.studocu+1

Verification ideas

1.    Match the test to the hypothesis. If you predict that CEO branding is positively associated with purchase intention or trust, use the right inferential test for the variable types and research design, such as correlation, t-test, ANOVA, or regression. A result is more believable when the test logic fits the data structure and the study question.studocu

2.    Check whether the result is statistically and practically meaningful. Do not rely only on the p-value; report effect size and confidence interval as well, because inferential statistics are about estimating population patterns from a sample, not just labeling a result as significant. A small but significant effect may be weak in practical MBA terms, while a moderate effect may matter more for branding strategy.studocu

3.    Look for internal consistency across measures. If several questionnaire items measure the same concept, the pattern should broadly agree across related items, subscales, and descriptive statistics. Consistency across indicators makes your conclusion more trustworthy than a single isolated significant result.hkuspace.hku+1

4.    Use robustness checks. Re-run the analysis with sensible alternatives, such as checking assumptions, comparing parametric and non-parametric tests, or testing models with and without control variables. If the conclusion remains similar, your finding is better verified.studocu

Falsification ideas

1.    State a clear null hypothesis that can fail. For example, “CEO branding has no relationship with brand trust among Hong Kong respondents” is falsifiable because the data can contradict it. In dissertation work, hypotheses should be precise enough that evidence can show they are unsupported.arxiv+1

2.    Actively look for disconfirming patterns. Test whether the relationship disappears in subgroups, such as by age, work experience, or gender, or whether the effect reverses after adding controls. Falsification is not about forcing a rejection, but about asking whether the claim still holds under harder tests.studocu

3.    Check for alternative explanations and bias. A result that seems to support your hypothesis may actually reflect sampling bias, social desirability bias, poor wording, or non-response bias. Hong Kong survey quality guidance emphasizes careful questionnaire design, high response rates, and thorough checking of raw data to avoid misleading conclusions.studocu

4.    Try to reproduce the pattern with a different specification. If the effect only appears with one coding choice, one subset, or one analytic model, the claim is weak. A finding that survives re-coding, re-grouping, or alternative model assumptions is harder to falsify, and therefore stronger.studocu

Applying it to your survey

For an MBA dissertation, the key is to present verification and falsification as part of a disciplined analysis process, not as abstract philosophy. You can write that verification comes from consistency, significance, effect size, and robustness, while falsification comes from null hypotheses, subgroup testing, bias checks, and alternative model comparisons. That framing shows you are not merely “proving” your idea about CEO branding, but testing it rigorously in a way that supports sound managerial conclusions.hkuspace.hku+1

A simple way to phrase this in your dissertation is: “The analysis sought both to verify the hypothesized association between CEO branding and respondent perceptions and to falsify the claim by examining alternative explanations, robustness across subgroups, and sensitivity to model specification”.

** also study the video on Popper and the logic of falsification.

** A collection of lecture notes on the subject of research methods for MBA students, 2026 June

A collection of blog notes on using chatgpt for research purpose.