Science fair project · Grades 6–8
Is popularity an equation?
This project studies how people think about friendship and social connections. It is not a contest and it does not rank real kids. You will test a simple math model, P = 2F + 3A − D², to see whether it lines up with how liked and respected someone feels. The numbers in the model are a guess to test—not a proof, and not a score of anyone’s worth.
The idea (and the algebra)
Someone wondered whether middle-school social life could be sketched with a simple equation. The model they proposed looks like this:
- P = the model’s predicted score (a number the equation invents—not something you measure directly)
- F = count of close trusted friends (people you can really count on)
- A = count of activities you plug into—sports, clubs, robotics, track, theater, and similar groups
- D = count of tense or stressful social moments in the last month (for example, arguments or rumors). This is a count of events, not a score that blames a person. A hard moment can happen to you without being your fault.
Why ×2, ×3, and D²?
In algebra, a coefficient is a multiplier in front of a variable. Here the coefficients are teaching tools and a hypothesis at the same time—guesses to test, which is the point of the project:
- Friends get a coefficient of 2 in the starting story: each close friend might also introduce you to about one more person. That story mixes friends with the people they introduce—it is a guess, not a measured fact.
- Activities get a coefficient of 3: the idea is that a team or club plugs you into a whole group at once. “A whole group” does not by itself justify 3 instead of 5 or 10—those numbers are placeholders to test.
- D is squared: D² means D × D. The small 2 (the exponent) tells you how many copies of D to multiply. So a count of 1 subtracts 1 point (1×1), but a count of 4 subtracts 16 points (4×4). The subtracted amount grows faster as the count gets larger—that is quadratic growth (the extra amount for one more is 2D+1). This is about the math, not about blaming a person.
What the model is saying
Under this guess, many activities and a low D count can outrank a high friend count that also has a high D count. That is an interesting claim. In a science fair, interesting claims belong as questions, not as facts.
Your project asks: when people report their own F, A, and D counts (safely and anonymously), or practice with a made-up character, does the predicted score P line up with the “liked and respected” rating? If not, can you find better relative weights from your own data?
Try the model (browser only)
Enter sample counts for friends (F), activities (A), and tense or stressful social moments (D). You can use made-up numbers or practice counts—not another real kid's name or score. The result is calculated in this browser only. Nothing is stored or sent. Use this to explore the model before you collect offline data.
P=2F+3A−D2
Predicted score P
11
6 + 6 − 1 = 11
Linear comparison
11
Same inputs with −D instead of −D² (identical when D is 0 or 1)
How D² grows
D² means D × D. The small 2 (the exponent) tells you how many copies of D to multiply. Going from D to D+1 subtracts 2D+1 more points (1st adds 1, 2nd adds 3, 3rd adds 5, 4th adds 7, 10th adds 19). That is quadratic growth. It is about the math, not about blaming a person.
| D | D² total | Extra cost of one more (2D+1) | |
|---|---|---|---|
| 0 | 0 | +1 | |
| 1 | 1 | +3 | |
| 2 | 4 | +5 | |
| 3 | 9 | +7 | |
| 4 | 16 | +9 | |
| 5 | 25 | +11 | |
| 6 | 36 | +13 | |
| 7 | 49 | +15 | |
| 8 | 64 | +17 |
Science fair steps
Question
Does the model P = 2F + 3A − D² predict a student’s self-rated “I feel liked and respected” score (1–10) better than chance—and better than a simpler linear model that only subtracts D once?
Hypothesis
If higher predicted scores P (more friends and activities, and a lower D count after squaring) also go with higher “I feel liked and respected” ratings, then the starting weights may be useful. If the pattern is weak or reversed, revise the model or the weights.
Variables
Predictor (explanatory) variables
In an experiment you would change these on purpose. Here nobody controls them—you only measure what people already report—so scientists call them predictors (or explanatory variables), not independent variables.
- F — self-reported count of close trusted friends
- A — self-reported count of ongoing activities/clubs/teams
- D — self-reported (or made-up character) count of tense or stressful social moments in the last month; no naming real people
Outcome (what you measure)
The only measured outcome is the Feeling rating. P is the model’s prediction—you compute it from F, A, and D; you do not survey people for P.
- Self-rated “I feel liked and respected” score from 1 (not at all) to 10 (very much)
Model prediction (not a measured outcome)
- P = 2F + 3A − D² — computed after the survey; use it on graphs as a predicted score to compare with Feeling
Controlled conditions
- Same anonymous survey wording for every participant
- Same time window for drama (“last month”)
- Same grade band (e.g. only 7th grade, or only your class)
- No names; no ranking of classmates; no “who is popular” lists
- Teacher and parent/guardian permission, student assent, and the right to skip or withdraw at any time with no grade penalty, before any data is collected
Method
- Get approval first. Many science fairs, including those following Regeneron ISEF rules, require approval from a school review committee (IRB/SRC) before you survey people. Ask your teacher before you start.
- Print the permission slip and data sheet. Collect parent/guardian consent, student assent, and teacher or school approval offline. Collect and store those forms separately from the anonymous response sheets.
- Aim for about 20–30 completed anonymous responses (not a tiny handful). More sheets make patterns clearer and leave room for a train/test split.
- Hand out anonymous paper surveys. Each person answers only about themselves, or about a made-up character you provide for practice. They fill in F, A, and D as counts, plus a 1–10 “I feel liked and respected” rating. No names on the response sheet. Never ask anyone to name, rank, or score a real classmate.
- Participation is voluntary. Anyone may skip a question or withdraw at any time, with no grade penalty.
- For each completed sheet, compute P = 2F + 3A − D² (use the calculator on this page or a spreadsheet offline).
- Optionally also compute P_linear = 2F + 3A − D for the extension comparison.
- Build a data table: Participant code (R01, R02…), F, A, D, D², P, P_linear, Feeling (1–10).
- Split the sheets into two piles (e.g. even codes vs odd codes, or first half vs second half). Fit relative weights on one pile; check whether they still work on the held-out pile (see Extensions).
- Graph Feeling vs P (scatter plot). Optionally add Feeling vs P_linear on a second plot.
- Compare average Feeling in the top-half P scores vs the bottom-half P scores (ranking/trend—do not try to make P equal Feeling, because the scales differ).
- Shuffle test (better than chance): write each Feeling score on an index card, shuffle, and re-deal the cards onto the fixed P values. Recompute the top-half vs bottom-half Feeling gap. Repeat 10–20 shuffles. Ask whether your real gap is bigger than almost all of the shuffled gaps.
- For squared vs linear: count how many responses have D ≥ 2. If almost everyone has D = 0 or 1, D² = D so the two models match—focus the comparison on sheets with D ≥ 2.
- Keep the paper responses with the student researcher and the teacher only. Discard them after the science fair (or sooner if your school requires it). Do not post answers online.
- Write analysis, conclusion, and limitations—including correlation vs causation and self-report bias.
Data table
Use codes (R01, R02…) instead of names. Print the blank sheet or copy these columns into a notebook:
| Code | F | A | D | D² | P | P_linear | Feeling 1–10 |
|---|---|---|---|---|---|---|---|
| R01 | — | — | — | — | — | — | — |
| R02 | — | — | — | — | — | — | — |
| R03 | — | — | — | — | — | — | — |
Graphs
Make a scatter plot with predicted score P on the x-axis and Feeling (1–10) on the y-axis. Look for a trend (higher P with higher Feeling)—not for P to equal Feeling, because the scales differ. Optionally add a second chart for P_linear. Label axes clearly.
Analysis
- Do higher P values tend to go with higher Feeling scores (top-half vs bottom-half averages, or pairs in the right order)? Remember P and Feeling use different scales—judge ranking/trend, not matching numbers.
- Was your real top-half vs bottom-half gap larger than almost all of the 10–20 shuffled gaps?
- How many responses had D ≥ 2? If few, can you fairly compare −D² with −D?
- If you fit P = F + bA − cD² on half the sheets, do those b and c still look useful on the other half?
- Are there outliers? What might explain them without guessing about real classmates?
Conclusion
- Was your hypothesis supported, partially supported, or not supported?
- What would you change about the model or the survey if you ran it again?
- What did you learn about using math as a model of social life—not a verdict on a person?
Limitations
- This is correlational, not causal. Even a strong pattern does not prove that joining a club causes feeling liked and respected.
- The project title talks about “popularity,” but the measured outcome is “I feel liked and respected.” Those are related ideas, not the same thing.
- Close friends (F) likely raise that feeling almost by definition, which can make the friend term look stronger than a fair popularity test would.
- A 1–10 rating is ordered categories (like placing yourself on steps), not a true equal-interval measurement scale—so averages are a rough summary.
- Self-report bias: people may miscount friends, activities, or stressful moments, or rate feelings differently on different days.
- The sample is not the whole school—only people who opted in with permission. Fitting many coefficients on a small sample can overfit; use 20–30 responses and a train/test split.
- D mixes different kinds of hard moments. A higher D count is not a judgment of a person, and it does not mean someone caused the problem.
- Coefficients 2 and 3 were invented as a starting hypothesis, not measured from a large study. The ×2 and ×3 stories are motivational guesses.
- Social life is richer than three variables. The model is intentionally simple so middle-school algebra can meet a fair test.
Extensions
Fit relative weights (friend coefficient fixed at 1)
P from the starting model can be negative or much larger than 10, while Feeling is always 1–10—so never judge fit by making P equal Feeling. Instead, ask whether higher P ranks with higher Feeling (top-half vs bottom-half averages, or how often pairs line up). Multiplying every coefficient by the same number does not change rankings, so fix the friend weight at 1 and try P = F + bA − cD² with small whole-number b and c. Fit on half your sheets; test on the other half. Report that the weights are fitted to your sample, not universal laws.
Squared drama vs linear drama
Compare P = 2F + 3A − D² with P_linear = 2F + 3A − D using ranking/trend against Feeling. Important: when D is 0 or 1, D² = D, so both models give the same score. Count how many responses have D ≥ 2; if most people have D ≤ 1, the comparison cannot tell the models apart—focus on the D ≥ 2 subset (or collect more of those responses).
Correlation vs causation discussion
Write a short paragraph: even if active kids report feeling more respected, clubs might attract kids who already feel connected—or a third factor (sleep, adult support, class climate) might drive both. Good science fair boards say what they cannot prove.
Safe, ethical data plan (offline only)
- All survey data is gathered offline on paper (or a school-approved offline tool). The calculator on this site runs only in the browser. Nothing you type there is stored or sent. This project has no online survey and no forms that collect personal information.
- Anonymous responses only. Use codes like R01, not names. Never write a real person’s name on a response sheet.
- Get teacher or school approval, parent/guardian permission, and the student’s own yes (assent) before inviting anyone. Collect and store those forms separately from the responses.
- Participation is voluntary. Anyone may skip a question or withdraw at any time, with no grade penalty.
- Each person answers only about themselves, or about a made-up character you provide for practice. They report counts of close friends, activities, and tense or stressful social moments in the last month, plus a private 1–10 “I feel liked and respected” rating.
- Never ask anyone to name, rank, or score a real classmate. There is no “who’s popular” list in this project.
- Paper responses stay with the student researcher and the teacher only. Discard them after the science fair (or sooner if your school requires it). Do not post answers online.
- Many science fairs, including those following Regeneron ISEF rules, require approval from a school review committee (IRB/SRC) before you survey people. Ask your teacher before you start.
A kind note
A person’s worth is not a number. The equation is a simple model for testing an idea—not a grade, a rank, or a verdict on anyone. If a question feels uncomfortable, skip it. If something at school or online is bothering you, talk to a trusted adult.
The calculator on this page runs only in your browser. It does not send or store your numbers. This project has no online survey and no forms that collect personal information.