Missing Data — Explained
Detailed Explanation
When can a blank entry be calculated?
A missing-data question asks whether the supplied relationships determine an unknown value. A blank is not automatically zero. It may be recoverable from a total, a ratio, an average or another independent constraint; it may also remain undetermined. Begin by identifying what each row and column represents and whether its categories overlap. A total can be used by subtraction only when the listed parts form the whole described by that total.
Use an original example: a library records 240 books issued on Monday, 310 on Tuesday and an unknown number on Wednesday. The three-day total is 900. Wednesday's count is 900 - 240 - 310 = 350. This is an exact recovery because the total covers precisely those three days. If the total instead covered a full week, subtracting the two known days would leave five days' combined count, not Wednesday's value.
How do row and column totals work together?
Consider a table of workshop registrations. Centre A has 40 morning participants and 25 afternoon participants. Centre B has 35 morning participants and an unknown afternoon count. If total registrations are 130 and each registration appears once, the unknown is 130 - 40 - 25 - 35 = 30. The morning total is 75 and the afternoon total is 55, providing a second way to check the completed table.
If both centres' afternoon cells were blank and only their combined total of 55 were supplied, each individual count would remain undetermined. The pairs 20 and 35, 25 and 30, or 30 and 25 all satisfy that constraint. One equation with two independent unknowns generally leaves freedom. A further fact, such as one count exceeding the other by five, could resolve it. More printed numbers do not necessarily mean more independent information.
How can an average reveal a missing value?
Convert an average back into a total. If five daily outputs average 46 units, their sum is 230. If four outputs are 38, 44, 51 and 47, their sum is 180, so the missing output is 50. The denominator five is essential: subtracting the four known values directly from the average mixes an individual-level measure with a total.
For group averages, use group sizes. A class of twenty students averages 60 marks and another class of thirty averages 70. Their combined average is (20 × 60 + 30 × 70)/50 = 66, not 65. If the combined average and one group's average were known, the same weighted-total equation could recover the other. Averaging group averages directly is justified only when the relevant group sizes are equal or the question defines a different weighting rule.
What should be done with ratios and percentages?
Translate the ratio into parts of a stated whole. If two categories are in the ratio 3:5 and together contain 160 items, one part equals 20, so the counts are 60 and 100. If a third category exists outside the stated total, it must not be included without evidence. Ratios describe relative quantities; their denominator is not necessarily the overall table total.
Suppose 40 per cent of 250 applicants qualify and 60 qualified applicants are from one group. The remaining qualified applicants number 100 - 60 = 40. Subtracting 60 from 250 would instead count everyone outside that named subgroup, including unqualified applicants. Draw the hierarchy—applicants, qualified applicants, named qualified subgroup—before subtracting. Nested categories are a common source of plausible but semantically wrong answers.
What if the table contains rounded percentages?
A rounded percentage may determine a range rather than a unique count. If an exact population of 1,000 has a reported share of 20 per cent rounded to the nearest whole percentage point under a stated half-up rule, the underlying share lies from 19.5 per cent inclusive to 20.5 per cent exclusive. Integer counts from 195 through 204 are consistent with that display. It is unjustified to assert exactly 200 from the rounded report alone.
Similarly, displayed percentages can sum to 99 or 101 because individual categories were rounded. Do not repair the table by assigning the discrepancy to a particular blank unless the question supplies the necessary convention or exact totals. Distinguish exact data, rounded data and estimates. A question asking for the nearest approximate value permits a different conclusion from one asking for the uniquely determined original count.
How are overlapping sets handled?
When two categories overlap, adding their counts double-counts the intersection. Suppose 80 people attend a history session, 65 attend an economics session and 30 attend both. The number attending at least one is 80 + 65 - 30 = 115. If the total population is 150, the number attending neither is 35. The arithmetic depends on knowing that these are the two sessions relevant to the classification.
If the question instead supplies the union and asks for the overlap, rearrange the same relationship: intersection = first count + second count - union. An overlap cannot exceed either individual set and cannot be negative. These bounds offer quick checks. Treating 'at least one', 'exactly one' and 'both' as interchangeable changes the quantity being counted even when all the numbers are copied correctly.
How should insufficient or inconsistent data be recognised?
Try constructing two different completions that satisfy every supplied condition. If both give different values for the requested cell, the data are insufficient. This is stronger than merely feeling that a number is missing. Conversely, if the proposed totals force a negative participant count, an overlap larger than its parent set or contradictory sums, the data or interpretation are inconsistent.
The final routine is to label the unknown, convert prose into equations, check independence and verify the completed values against the original table. Keep units and periods aligned: monthly totals do not combine directly with annual averages, and percentages do not become counts without a denominator. A good solution explains both the arithmetic and why the available information is sufficient to support that arithmetic.