Punggol Secondary 3 English tuition should help students notice a hidden weakness in persuasive evidence: sometimes the examples we see are not representative of the cases that existed. Selection bias occurs when the observed sample is chosen in a way that distorts the conclusion; survivorship bias is a special case where failures disappear from view and only survivors remain.
This rebuilt 2019 legacy URL now owns one distinct Sec 3 English job: claim → visible sample → selection process → missing cases → bias direction → repaired conclusion. Old mixed-grade commercial copy and legacy media have been removed.
eduKateSG’s current Punggol centre is at 83 Punggol Central, Singapore 828761. Selected Secondary English classes run in focused 3-pax small groups, typically 1.5 hours.
The Direct Answer
Selection bias asks:
“Who or what entered the evidence set—and who did not?”
Survivorship bias asks:
“Are we studying only the cases that survived, succeeded or remained visible?”
The evidence you can see may be systematically different from the evidence that disappeared.
Worked Example: Study Advice
Article interviews ten top scorers about their study routines.
Conclusion:
“Top scorers wake at 5 a.m.; therefore waking at 5 causes strong results.”
Problem:
- only successful students were sampled;
- students who woke early but performed poorly are missing;
- other factors may matter.
Selection Bias
Suppose a survey asks:
“Do you enjoy reading?”
but recruits respondents from the library club.
The sample likely overrepresents enthusiastic readers.
A conclusion about the whole school would be biased.
The Sampling-Frame Test
Ask:
From which group were respondents actually selected?
Target population:
all Secondary 3 students.
Sampling frame:
students attending a book fair.
Mismatch creates risk.
Volunteer Bias
Online poll:
“Tell us how angry you are about the new policy.”
People with strong feelings may be more likely to respond.
The sample of volunteers may not represent quiet/neutral people.
Worked Example: Product Reviews
Only customers motivated enough to post reviews are visible.
Possible missing groups:
- satisfied but silent users;
- people who returned product without reviewing;
- people who never bought it.
Review averages can still be useful, but the selection process matters.
Survivorship Bias
Imagine studying only businesses that survived ten years.
Shared trait:
their founders took large risks.
Conclusion:
“Large risk-taking causes business survival.”
Missing:
businesses that took equally large risks and failed.
The Missing-Failures Test
Whenever success stories are used as evidence, ask:
How many people used the same strategy and did not succeed?
Without failures, success factors are hard to identify.
Worked Example: Exam Testimonials
Advertisement:
“These five students improved by two grades using Method X.”
Questions:
- how many students used Method X in total?
- what happened to the rest?
- were these five selected because they improved most?
- what was their starting point?
This is why reader-safe tuition writing avoids unsupported outcome promises.
Selection on the Outcome
If researchers select only successful cases and then search for common causes, they cannot tell whether the same features also existed among failures.
Outcome-based selection can manufacture patterns.
Worked Example: Sports
Study only professional athletes.
Find that many trained intensively as teenagers.
Cannot conclude:
intensive teenage training guarantees professional success.
Many non-professionals may have trained equally hard.
Availability Bias vs Selection Bias
A dramatic event may be easier to remember.
That is an availability problem.
Selection bias concerns how cases entered the dataset.
The ideas can interact but are distinct.
Publication Bias
Studies with exciting positive findings may be more visible than studies showing no effect.
If readers see only published positive results, they may overestimate the strength of a phenomenon.
At Sec 3, students need the basic idea: missing neutral/negative results can distort the visible evidence.
Worked Example: “Nine Successful Startups”
Book analyses nine successful startups and lists:
- founder obsession;
- rapid growth;
- long hours.
Before treating these as causal success rules, ask whether failed startups shared the same traits.
Selection Bias in News
Headline:
“Residents furious about new scheme.”
Article interviews four angry residents.
Questions:
- how were interviewees chosen?
- were supportive/neutral residents approached?
- does the article claim all residents or only some?
Anecdotal Selection
A writer may choose the most dramatic example because it is memorable.
One tragic case can be important, but it may not reveal typical frequency or probability.
Use anecdote for human impact; use representative evidence for prevalence.
Worked Example: School Discipline
Argument:
“Phones are always disruptive. I saw three students using them during lessons.”
Missing denominator:
How many students used phones appropriately or did not use them?
Selection may focus on violations because violations attract attention.
The Denominator Question
Visible cases:
10 complaints.
Missing information:
10 complaints out of:
- 20 users?
- 20,000 users?
Selection analysis often pairs with denominator/base-rate reasoning.
Worked Example: Successful Revision Videos
Students who post “I used this technique and scored A” are visible.
Students who used it and did not improve may not post.
Testimonials alone cannot establish effectiveness.
Selection Bias Through Exclusion Rules
Survey excludes students absent on the survey day.
If absenteeism relates to the topic, the exclusion may bias results.
Always inspect who was eligible.
Selection Bias Through Access
Online-only survey may miss people with weak internet access.
Phone survey may miss people who do not answer unknown numbers.
Mode of collection changes who appears.
The Missing-Case Inventory
For any evidence set, list:
- included successes;
- included failures;
- excluded successes;
- excluded failures;
- silent/non-responding cases.
Even a rough inventory reveals blind spots.
Direction of Bias
Ask:
Would the missing cases likely make the result look stronger, weaker or simply more uncertain?
Library-club sample likely overstates reading enthusiasm.
Complaint forum likely overstates dissatisfaction.
Not every selection bias direction is obvious, so avoid guessing.
Worked Example: Employee Survey
Only current employees are surveyed about workplace culture.
Former employees who resigned because of poor culture are missing.
This can create survivorship bias.
Worked Example: Historical Buildings
If we study old wooden houses still standing today, we may conclude that old wooden construction was extremely durable.
But the houses that decayed, burned or were demolished are missing.
Survivors are a selected subset.
Selection Bias vs Cherry-Picking
Selection bias may arise from how data is collected.
Cherry-picking often means a writer selectively presents favourable data from a broader set.
Both create unrepresentative evidence.
Worked Example: Time Window
Policy evaluation uses only the best-performing three months.
If poorer months are omitted, the selected timeframe can bias the conclusion.
Selection Bias in Comparisons
Compare:
top 10 students in School A
with
all students in School B.
Even if the metrics match, selection rules differ.
Like-for-like comparison requires comparable sampling.
The Representativeness Test
Ask:
- Who is the target population?
- Who could enter the sample?
- Who actually entered?
- Who was excluded/silent?
- Does inclusion relate to the outcome?
Repairing the Evidence
Possible repairs:
- sample randomly/stratify;
- include failures/non-responders;
- report selection criteria;
- compare included vs excluded groups;
- narrow the conclusion to the observed sample.
Narrow Claim Repair
Too broad:
“Students prefer digital notes.”
Data from online study group only.
Repair:
“Most respondents in this online study group preferred digital notes.”
Still useful, but honest about population.
Selection Bias in Comprehension
If a writer says:
“Everyone we interviewed agreed…”
ask:
Who was interviewed?
The statement may accurately describe the sample while misleading readers about the population.
Selection Bias in Argumentative Writing
Before citing:
“Many successful people…”
ask whether you are ignoring unsuccessful people who used the same behaviour.
This protects essays from inspirational anecdote masquerading as causal evidence.
The Visibility Test
Ask:
What kinds of cases naturally become visible?
Successful creators get interviews.
Failed attempts disappear.
Complaints appear on complaint pages.
Satisfied users may remain silent.
The Bias Audit
| Question | Purpose |
|---|---|
| Target population? | define who claim covers |
| Selection rule? | see who could enter |
| Missing cases? | find blind spots |
| Outcome affects visibility? | detect survivorship |
| Bias direction? | calibrate conclusion |
Sec 3 Context
MOE’s Secondary English Language Syllabus 2020 emphasises critical evaluation of evidence and persuasive texts. Selection and survivorship bias give Sec 3 students a concrete way to ask not only “What evidence is shown?” but “What evidence might be missing?” Official reference: MOE Secondary English Language Syllabus 2020.
Common Selection-Bias Failure Modes
| Symptom | Problem | Repair |
|---|---|---|
| success stories treated as causal proof | failures invisible | seek non-survivors |
| club survey generalised to school | sample frame narrow | limit population |
| voluntary poll treated representative | self-selection | check non-responders |
| headline interviews only angry residents | selection unclear | ask recruitment method |
| best months selected | time cherry-picking | use full relevant period |
Why Three Students Helps
One student defends the visible evidence, one lists missing cases, and the third decides how much the claim must narrow once the selection process is exposed.
What Progress Should Look Like
A stronger Sec 3 learner can identify the target population, inspect how cases entered the sample, recognise missing failures/silent groups and calibrate conclusions to the evidence that was actually observable.
Current class enquiries: selected eduKateSG Secondary 3 English 3-pax small groups, typically 1.5 hours. Punggol centre: 83 Punggol Central, Singapore 828761. Contact +65 8823 1234.