The Mall Sample
A research team is studying household spending on packaged food in a Tier-2 city. With a two-week deadline and a budget for 300 interviews, they station surveyors at two upscale malls on weekends. They obtain 280 responses; 62% of respondents report monthly household income above ₹80,000, whereas the city's latest official estimate puts that bracket at about 18% of households. The draft claims the findings represent 'urban middle-class consumption patterns' and recommends a statewide marketing strategy. A team member suggests weighting the data post hoc using income quotas; another suggests adding a caveat. The funder wants a headline figure within five days. The questionnaire also asks about monthly spending, but 34 respondents left it blank and were excluded from the analysis, and the draft reports the mean for the remaining 246 without mentioning the exclusions.
1. Evaluate the sampling design and the claim that the findings represent urban middle-class consumption patterns.
Read the worked answer
The design is a convenience sample of weekend shoppers at two upscale malls, not a probability sample of households. The income skew is diagnostic: 62% of respondents above ₹80,000 versus an official estimate of about 18% of households shows strong selection bias. Post hoc income weighting can partially adjust estimates, but it cannot repair the sampling frame, which excludes non-mall shoppers, weekday shoppers, and lower-income households. The claim of 'urban middle-class consumption patterns' is therefore overreach; the defensible claim is 'among weekend shoppers at two upscale malls in this city.' A statewide marketing recommendation is unsupported by such a sample. The 34 missing spending values are also serious: if nonresponse is related to income or spending, complete-case analysis of 246 can bias the mean. The draft must report the response rate, exclusions and limitations.
2. What should the team do about the missing spending data and the weighting suggestion?
Read the worked answer
First, report the 34 exclusions transparently and compare the income profile of respondents who answered and did not answer the spending question. If the missingness appears related to income, complete-case analysis is suspect. The team can use multiple imputation or a sensitivity analysis that presents a range of plausible means, and it should report the unweighted and weighted estimates side by side. Post hoc income weighting is not automatically wrong, but it must be pre-specified in the analysis plan, based on a credible external distribution, and reported with its assumptions. Weighting on a single variable cannot correct for mall-based selection. The team should also avoid presenting a single headline mean as if it were a population estimate; confidence intervals and a clear limitation statement are necessary.
3. How should the team respond to the funder's demand for a headline figure within five days?
Read the worked answer
The team should give the funder a defensible headline, not a false one: for example, 'among 246 weekend mall shoppers who reported spending, mean monthly packaged-food spending was ₹X.' It should attach the caveats about the sample and missing data, and explain why a statewide recommendation cannot be made. If the funder needs a population estimate, the team should propose a small probability-based or quota sample in lower-income areas, even if that means a later deadline. The five-day deadline does not justify relabelling a convenience sample as representative. The team should document the funder's request and its response, because pressure to overstate findings is itself a research-integrity issue. A short, honest interim report is preferable to a precise but misleading figure.