Mathematics: Applications & Interpretation SL
Mathematical exploration preparation: data and model critique
- 1.
In an exploration-preparation exercise, hypothetical pairs (x,y) are (1,3),(2,4),(3,8), where x is hours and y is a dimensionless score. Fit the least-squares line y=a+bx using b=Sxy/Sxx and a=ybar-b xbar. Calculate the residuals observed minus fitted and their sum. Explain why a zero residual sum does not prove the line is a good model.
[6 marks]Worked explanation
- Centre the x values at 2 and the y values at 5. Sxx=1+0+1=2; Sxy=(-1)(-2)+0(-1)+1(3)=5.
- The slope is 5/2=2.5 and intercept is 5-2.5(2)=0. Therefore fitted y values are 2.5,5,7.5.
- Subtract fitted from observed: 0.5,-1,0.5. They sum to zero but their squared sum is 1.5, not zero.
- Residual balance is a fitting property, not a goodness-of-fit certificate. Three observations are too few to establish a stable trend; additional independent data and residual checks would be useful.
Common mistake: Zero signed residual sum allows positive and negative errors to cancel; it does not mean every prediction is exact.
Marking points
- Finds xbar=2, ybar=5, Sxx=2 and Sxy=5.
- Obtains b=2.5 score units per hour and a=0.
- Calculates fitted values 2.5,5,7.5.
- Calculates residuals 0.5,-1,0.5.
- Finds residual sum 0.
- Explains that least-squares fitting with an intercept makes residuals sum to zero even when curvature or other misfit exists; the positive-negative-positive pattern warrants further checks.
Examiner tip: Zero signed residual sum allows positive and negative errors to cancel; it does not mean every prediction is exact.
- 2.
For exploration preparation, a hypothetical account starts with $2500 and has no deposits, withdrawals or fees. Model its balance after 10 years at annual compound rates of 4% and 5%. Calculate both balances and their difference to the nearest cent, retaining unrounded values until the end. Explain why this comparison is a sensitivity analysis rather than evidence of future returns, and identify an omitted financial factor.
[5 marks]Worked explanation
- Annual compounding multiplies by 1+r ten times: 2500(1.04)^10=3700.610712... and 2500(1.05)^10=4072.236567... .
- The unrounded difference is 371.625854..., so the balances are $3700.61 and $4072.24 and the difference is $371.63.
- This varies the assumed rate while fixing other inputs. It shows a consequence of assumptions, not an investment forecast.
- Inflation is absent: even a larger nominal balance may have less purchasing power than these dollar figures suggest. Fees or variable annual rates would also change the model.
Common mistake: Adding 10 times one year's interest uses simple interest and misses interest earned on earlier interest.
Marking points
- Uses B=2500(1+r)^10.
- Obtains $3700.61 at 4%.
- Obtains $4072.24 at 5% and difference $371.63 from unrounded values.
- Explains that changing an assumed input tests model sensitivity; neither constant rate is an observed or guaranteed future return.
- Identifies an omitted factor and its effect, such as inflation lowering the purchasing power of a nominal balance.
Examiner tip: Adding 10 times one year's interest uses simple interest and misses interest earned on earlier interest.
- 3.
An exploration-preparation table contains hypothetical travel times from volunteers: 20 bus users have mean 30 minutes and 80 walkers have mean 12 minutes. Calculate the mean for all 100 volunteers. Explain why averaging 30 and 12 is wrong for this sample. A writer concludes that every student at the school averages this travel time. Give one sampling limitation and a specific improvement to the proposed collection plan; do not claim that new data have been collected.
[5 marks]Worked explanation
- Convert each group mean back to a total: 600 bus-user minutes and 960 walker minutes. The combined total is 1560 minutes for 100 people.
- The sample mean is 15.6 minutes. Equal averaging gives 21 minutes and would describe equal-sized groups, not the stated 20:80 split.
- Volunteers may differ systematically from other students, and car or cycle users are not represented. The result describes only the hypothetical sample.
- A defensible next plan draws students randomly from the whole school and addresses nonresponse; proportional stratification can include each travel mode. This is a proposal, not completed data collection.
Common mistake: A larger sample size does not by itself remove bias from the way volunteers were selected.
Marking points
- Calculates total hypothetical travel time 20(30)+80(12)=1560 minutes.
- Obtains overall sample mean 1560/100=15.6 minutes.
- Explains that the two groups have unequal sizes; (30+12)/2=21 minutes incorrectly weights them equally.
- Identifies voluntary-response bias or exclusion of other travel modes as limiting school-wide inference.
- Proposes randomly sampling from a school-wide list, optionally stratified by travel mode with proportional weighting, rather than claiming the volunteer result represents everyone.
Examiner tip: A larger sample size does not by itself remove bias from the way volunteers were selected.
- 4.
For written exploration preparation, a hypothetical tank model is V(t)=120-8t litres, where t is minutes after draining starts. It was fitted using observations only from t=2 to t=8. Calculate V(5) and V(20), classify each prediction as interpolation or extrapolation, and give the physically permissible time domain if this formula describes draining from a full tank. Explain why a permissible prediction is not necessarily a validated one.
[5 marks]Syllabus placement evidence: Mathematics AI guide, first assessment 2021: SL 2.5 modelling; internal assessment, Criterion D: Reflection. Applies model-domain restrictions and evaluates predictions beyond fitted observations. This is AI SL exploration preparation using invented inputs, not an authenticated exploration or an IA criterion award.
Worked explanation
- Substitution gives 120-8(5)=80 and 120-8(20)=-40. Only 5 lies within the fitting interval [2,8].
- The tank empties when 8t=120. Restrict time to [0,15] for this model; a real flow rate may change with water depth, so domain restrictions are necessary but not sufficient evidence of accuracy.
Common mistake: A formula can return a number outside its physical domain; that does not make the number a possible tank volume.
Marking points
- V(5)=80 litres, an interpolation.
- V(20)=-40 litres, an extrapolation.
- Solves V=0 to obtain t=15 minutes.
- Physical domain is 0 <= t <= 15; negative water volume is impossible.
- Even within that domain, a constant draining rate needs evidence; physical plausibility alone does not validate the model.
Examiner tip: A formula can return a number outside its physical domain; that does not make the number a possible tank volume.
- 5.
An exploration-preparation dataset contains hypothetical waiting times 4, 5, 5, 6 and 30 minutes. Calculate its mean and median. Recalculate both after temporarily excluding 30 for a sensitivity check. Explain what the comparison shows and specify what evidence would be needed before permanently excluding that observation. No data-entry error is established.
[4 marks]Syllabus placement evidence: Mathematics AI guide, first assessment 2021: SL 4.3 measures of central tendency; internal assessment, Criterion D: Reflection. Uses mean/median calculations to reflect on sensitivity and justified inclusion. Written hypothetical preparation does not establish personal engagement, real data collection or a completed assessed exploration.
Worked explanation
- The central ordered observation is 5, whereas the large final observation raises the total to 50. Means and medians answer different descriptive questions.
- The temporary calculation isolates influence, not legitimacy. A documented transcription or measurement failure could justify correction/exclusion, but an inconvenient genuine event remains relevant to the defined population.
Common mistake: An influential observation is not automatically an erroneous observation, and sensitivity analysis is not permission to delete it.
Marking points
- Original mean=50/5=10 minutes and median=5 minutes.
- Without 30, mean=20/4=5 minutes and median=(5+5)/2=5 minutes.
- The mean changes by 5 minutes while the median is unchanged, showing different sensitivity to the extreme value.
- Check the original record and a predeclared inclusion rule; a genuine long wait should not be removed merely to improve a summary.
Examiner tip: An influential observation is not automatically an erroneous observation, and sensitivity analysis is not permission to delete it.
- 6.
For written spatial-model preparation, the only three monitoring sites in a rectangular region 0<=x<=8, 0<=y<=6 are A(0,0), B(8,0), C(0,6). Coordinates are in kilometres. Hypothetical rainfall readings are A:12 mm, B:20 mm, C:8 mm. Find the perpendicular bisectors of AB and AC, their intersection V, and the distance from V to each site. Determine the nearest site to P(2,4) by comparing squared distances, and use nearest-neighbour interpolation to estimate rainfall at P. Explain why a unique rainfall estimate at V requires a tie-breaking convention.
[7 marks]Syllabus placement evidence: Mathematics AI guide, first assessment 2021, August 2025 update: SL 3.5 and SL 3.6, printed p.44 (PDF p.50). Voronoi work is absent from the inspected AI SL tasks. The August 2025 edition confirms this bounded placement. Spatial model preparation in an existing broad topic only; no completed exploration is claimed.
Worked explanation
- AB is horizontal with midpoint (4,0), giving vertical bisector x=4. AC is vertical with midpoint (0,3), giving horizontal bisector y=3.
- At V(4,3), squared distances to the three sites are 16+9,16+9 and 16+9. Each distance is 5 km. Since there are no other sites, this is a genuine common cell vertex.
- At P(2,4), compare 2^2+4^2=20, (-6)^2+4^2=52 and 2^2+(-2)^2=8. The smallest identifies C; the model assigns C's 8 mm to P.
- V is on the three-way boundary. Equal distances do not imply equal rainfall readings, and taking their mean would add a rule not contained in nearest-neighbour interpolation. These hypothetical readings are not collected observations.
Common mistake: A complete perpendicular-bisector line need not be a Voronoi edge everywhere; compare against every site before claiming nearest-site membership.
Marking points
- The perpendicular bisector of AB is x=4.
- The perpendicular bisector of AC is y=3.
- Their intersection is V(4,3), within the specified region.
- VA=VB=VC=5 km, so V is a three-site Voronoi vertex.
- At P, squared distances to A,B,C are respectively 20,52,8 km^2; C is uniquely nearest.
- Nearest-neighbour interpolation gives 8 mm at P, using the reading at C.
- All three sites are equally close to V and have different readings; this nearest-site rule alone does not choose a unique value at V. A stated tie convention is needed.
Examiner tip: A complete perpendicular-bisector line need not be a Voronoi edge everywhere; compare against every site before claiming nearest-site membership.
- 7.
For written spatial-model preparation, a hypothetical narrow strip has perpendicular widths 12,18,21,17,10 metres at stations x=0,5,10,15,20 metres respectively. Use the equal-width trapezoidal rule to estimate its area. A separate finer-grid calculation gives 329 m^2. Find coarse estimate minus finer estimate, then this difference as a percentage of 329. Explain why the comparison alone cannot determine the true area or prove a particular error bound.
[5 marks]Syllabus placement evidence: Mathematics AI guide, first assessment 2021, August 2025 update: SL 5.8, printed p.66. Adds the previously absent AI SL equal-width trapezoidal calculation from a table. The lake/strip interpretation uses supplied widths and does not require an analytic boundary function.
Worked explanation
- Each neighbouring pair of widths forms a trapezium of area 5 times their average. Summing gives 75+97.5+95+67.5=335 m^2, an independent check of the weighted formula.
- Only the three interior widths are doubled. The four spaces between five stations set h; do not divide the total length by five.
- The finer estimate is lower by 6 m^2, not an established exact benchmark. Unsampled boundary curvature or measurement uncertainty can still matter; the readings here are hypothetical.
Common mistake: The first and last widths have half weight, and a finer approximation is not automatically the exact area.
Marking points
- Identifies four intervals of width h=5 m and five width measurements.
- Sets up area approximately (5/2)[12+2(18+21+17)+10].
- Obtains a coarse estimate of 335 m^2.
- Calculates difference 335-329=+6 m^2 and relative difference (6/329) times 100%=1.82% to three significant figures.
- Explains that both values are approximations; agreement or refinement without further information about the boundary is not proof of the true area or a guaranteed error bound.
Examiner tip: The first and last widths have half weight, and a finer approximation is not automatically the exact area.
- 8.
For written facility-location preparation, original sites are A(0,0), B(8,0), C(0,6), D(8,6), in kilometres. Only three locations are permitted for a hypothetical facility: P(4,3), Q(2,3), R(6,3). Its clearance is the distance to the nearest site; maximise this minimum distance over these candidates only. Calculate all three clearances and select the best before and after adding site E(4,3). Find the perpendicular-bisector line of A and E and explain how its E-side can restrict A's old Voronoi cell. Do not claim an optimum over every point of the plane.
[6 marks]Syllabus placement evidence: Mathematics AI guide, first assessment 2021, August 2025 update: SL 3.5-3.6, printed p.44. Adds a site and evaluates changed cells and clearances, beyond the first reassessment's static nearest-neighbour task. The maximin criterion and finite candidate set are explicit; this does not claim complete toxic-waste-location examination coverage.
Worked explanation
- For P, every old squared distance is 25. For Q the smallest is 2^2+3^2=13 at A and C; R has the same value at B and D. Compare minima, not the largest distance to a site.
- Distances to E at the candidates are 0,2,2. Each updated clearance is the smaller of the old clearance and the new E-distance, so the best candidate changes.
- Equate x^2+y^2=(x-4)^2+(y-3)^2 to get 8x+6y=25. The E-distance squared minus the A-distance squared is 25-8x-6y, negative on E's side.
- A bisector is an equidistance line, not necessarily an entire Voronoi edge. The finite candidate restriction is part of this original planning task; no global continuous-location or exam-style three-edge-intersection optimum is asserted.
Common mistake: Adding a site cannot increase the nearest-site distance at a fixed point; maximise the minima only over the permitted candidates.
Marking points
- Before adding E, finds clearance(P)=5 km.
- Before adding E, finds clearance(Q)=clearance(R)=sqrt(13) km; hence P alone is best among the three.
- After adding E, finds clearances 0,2,2 km at P,Q,R respectively.
- Chooses Q and R as equally best permitted locations after the addition, with clearance 2 km.
- Finds the A/E equidistance line 8x+6y=25.
- Explains that 8x+6y>25 means closer to E than A, so any such points previously in A's cell leave it; the full new cell must also satisfy comparisons with B,C,D.
Examiner tip: Adding a site cannot increase the nearest-site distance at a fixed point; maximise the minima only over the permitted candidates.