Eighteen Days, Six Hundred and Forty Students
Ananya, a Class XII student in a Kolkata school, has 18 days to complete her WBCHSE project on 'smartphone use at night and daytime sleepiness' among students of Classes XI and XII. The school has 640 such students across Science, Commerce and Arts. Her budget is ₹1,500; she has one teacher-supervisor who insists that no respondent be identifiable; the vice-principal wants 'a quick general finding about teenagers and phones'. Her first plan is to stand at the school gate at 3 p.m. and hand a questionnaire to whoever is willing; she gets 40 filled forms, though 12 of them leave the sleep-duration item blank and several of the rest give identical round numbers. Three girls have written 'my mother will scold me' next to a question about late-night chatting. Ananya must submit a method section, a table and a 300-word interpretation in 18 days.
1. Assess Ananya's gate-side plan. Name the sampling technique she has actually used, identify its main bias in this setting, and propose a feasible alternative design with numbers given her 18 days and ₹1,500.
Read the worked answer
What she has done. Standing at one gate at 3 p.m. and taking whoever agrees is a convenience (haphazard) sample, not a random one. It over-represents students who use that gate at that time and who are willing to discuss phones - the sociable, the unhurried, those without a 3 p.m. tuition or a bus to catch. Forty out of 640 is 6.25 per cent, and the 12 blanks show that even within that group self-selection continues to operate.
A feasible alternative. Use the class-section registers as a sampling frame (all 640 named). Stratify by stream, because phone and sleep habits plausibly differ between Science, Commerce and Arts, then draw a proportional random sample of about 80 with a random-number table - 12.5 per cent of each stratum, so roughly one in eight names per section. Print 90 forms to allow for absence, and use one class period per section; that fits inside 18 days and ₹1,500 easily.
Reporting. State the frame, the stratification, the achieved sample (say 76) and the absence/refusal rate. A small probability sample with a stated method supports a cautious generalisation to the 640; a larger convenience sample supports none. She must not describe 40 gate responses as 'the students of the school'.
2. Rewrite the question 'Why do you waste time on the phone at night?' and explain two problems with the original, including social desirability and how the supervisor's anonymity promise bears on the sleep item.
Read the worked answer
Problems with the original. (i) Leading or loaded wording: 'waste' carries a judgement, so respondents justify themselves instead of reporting behaviour. (ii) Vague and double-barrelled: 'at night' could mean 9 p.m. or 2 a.m., and 'why' invites rationalisation after the fact. (iii) Social desirability: students know adults disapprove of late-night phone use, so answers get 'improved' in the direction the question already signals.
Rewrite as neutral, factual items asked before any opinion item:
- 'On school nights last week, at what time did you usually stop using your phone? (write the time)'
- 'About how many minutes did you spend on the phone after 10 p.m. on a school night last week?'
Anonymity and the sleep item. The supervisor's promise means no names on the form. Students who fear a parent seeing the sheet will under-report late-night use - the three comments ('my mother will scold me') are direct evidence of that pressure. She should add a reassurance line, provide a sealed drop box, and pilot the form with five students to catch confusing items before printing 90.
3. How should she handle the 12 missing responses to the sleep item? Explain what she must and must not do, and how the missing data affect her interpretation.
Read the worked answer
Treat them as item non-response and report them. Twelve out of 40 is 30 per cent, which must appear in the method section and in the table, not be quietly dropped.
What she must not do: invent or impute a sleep figure for a respondent who left the item blank, or delete those 12 people from the whole analysis when they answered every other item.
What she should do first: check the pattern. If the blanks cluster in one class or stream, or come from students who also reported very late phone use, the missingness is not random and the sleep average will be biased - probably downwards, since heavy users may be the ones avoiding the item. If the pattern is scattered, the bias is smaller but still worth stating.
Practical fixes inside the deadline. A two-minute follow-up with the 12, identified by class and roll number rather than by name, is compatible with anonymity. If that is impossible, restrict the sleep finding to the 28 who answered and label it as such.
Interpretation. The 300-word report should say plainly that the sleep estimate is uncertain because a quarter of the forms were incomplete. Missing data is a finding about the difficulty of the topic, not a nuisance to be tidied away.