Final revision: a four-pass DI routine
First pass: identify the display type and read the title, legend, units and any total. In a table, trace the correct row and column; in a grouped bar chart, choose the correct colour for the required series; in a pie chart, write down the total for that pie. If the exact value is printed beside a bar or point, use it instead of judging its height visually. If the question asks for a value omitted from a chart, find the explicit total or relationship that determines it.
Translate the requested operation
Second pass: rewrite the question as a short expression before calculating. “Together” normally means addition, “difference” means subtraction, “how many times” means division, and “percent more” means difference divided by the stated comparison base. In a time series, a change from Year 1 to Year 5 uses those two endpoints; a total over five years uses all five points. When a chart has two axes, check the unit attached to each series before putting values into one expression.
Estimate, compute and check
Third pass: make a rough estimate. It should tell you whether the answer is in tens, hundreds or thousands, and whether a percentage is closer to 10%, 50% or 100%. Then perform the exact calculation. Fourth pass: compare your result with the options and check the unit. A count cannot be answered by a percentage, and a sector angle is measured in degrees, not units. If your exact answer is not listed, first recheck the values copied from the chart, then the denominator and any conversion. Do not force the nearest option when a chart value was misread.
A worked final check
Suppose a centre registered 250 candidates, 210 appeared and 170 qualified. The number who did not appear is 250−210=40. The number who appeared but did not qualify is 210−170=40. Although both answers happen to be 40, the groups are different. The qualification rate among those who appeared is 170/210, about 81%; among all registered candidates it is 170/250=68%. The question wording determines the denominator. This is why a correct-looking number is not enough: the working must match the named population.
Your error log after each set
For every missed item, record its ID and one cause: wrong chart element, wrong unit, wrong denominator, premature rounding, arithmetic slip or unsupported inference. Re-solve it without the printed answer. After one day, repeat the missed items; after one week, attempt a mixed set across tables, bars, lines, pies and caselets. A chart question is mastered only when you can name the data source and calculation before reading its explanation. These practice datasets are synthetic; their role is to train method, not to supply current public statistics.
Diagnostic worked case: table to comparison
A two-period table shows Unit A at 80 and 100, Unit B at 70 and 95, and Unit C at 60 and 90. First calculate the totals: Period 1 is 80+70+60=210; Period 2 is 100+95+90=285. The aggregate increase is 75 units. If the question asks for aggregate percentage growth, use the old total as the base: 75/210, about 35.7%. Do not average the three separate percentage growth rates, because their starting values differ. Unit A grows by 20/80=25%; Unit B by 25/70, about 35.7%; Unit C by 30/60=50%. The unweighted mean of these three rates is about 36.9%, which does not equal growth of the combined total. The table permits both questions, but their denominators differ. An efficient check is to compare the new total with 1.35 times the old total: 1.35×210=283.5, close to 285, so about 35.7% is sensible. If an option says 75%, it probably uses the 75-unit increase as if it were a percentage.
Diagnostic worked case: grouped bars and a histogram
A grouped bar chart shows blue and teal values for three units: A 40 and 60, B 70 and 80, C 90 and 75. The combined teal total is 60+80+75=215; the combined blue total is 200. Teal exceeds blue overall by 15, even though blue exceeds teal at Unit C. A question about one category cannot be answered from the overall totals. At B, teal exceeds blue by 10, which is 10/70≈14.3% relative to blue. Now switch to an equal-width histogram whose interval frequencies are 8, 12, 15, 10 and 5. Its total is 50 observations, and the most frequent interval is the third one. The histogram bars are adjacent because their horizontal positions represent connected intervals. The grouped bars are separated by named categories and colours denote series. Treating an interval label as an exact observed value would invent data. One may estimate a mean with midpoints if the task allows, but a count below the second upper boundary is exact: 8+12=20. This distinction tells you when reading an image gives a precise answer and when it only supports an estimate.
Diagnostic worked case: line trend and base change
A line records 50, 65, 60, 90 and 80 over five years. The first-to-last change is +30, so the overall percentage increase is 30/50=60%. The greatest adjacent rise occurs from Year 3 to Year 4: 90−60=30. The greatest adjacent fall occurs from Year 4 to Year 5: 90−80=10. The same number 30 appears in two different calculations; their meanings are different. If the question asks how much lower the final value is than the peak, compute 90−80=10 and use 90 as the percentage base if a percentage fall is requested: 10/90≈11.1%. Using 80 as base would answer by what percentage the peak exceeds the final value, 12.5%. The five-year total is 345, so the average is 69. To verify, the average lies between the minimum 50 and maximum 90 and is near the middle of the plotted points. A value outside that range would signal an addition or division error. Read the point labels, not the line slope alone.
Diagnostic worked case: pie share and actual count
Two pie charts describe different totals. In the first, Unit A is 20% of 400, or 80 units. In the second, Unit A is 15% of 600, or 90 units. Its share fell by five percentage points while its count rose by ten. The relative fall in share is 5/20=25%; the relative rise in count is 10/80=12.5%. All four statements can be true at once because the two wholes differ. If a question gives a sector angle instead, convert it by angle/360. An angle of 72° is one fifth of a whole, or 20%; with a total of 400 it represents 80 units. Check whether the total covers all sectors. If one chart reports only selected categories, the visible percentages might not sum to 100 and the missing portion cannot simply be called a named group without evidence. In the practice charts every share and total is printed, so work from those exact values. A visual slice that looks larger is not enough when two pies have different totals or different sizes on the page.
Diagnostic worked case: caselet and missing value
A selection caselet gives three centres. A registered 300, appeared 270 and qualified 180. B registered 250, appeared 225 and qualified 150. C registered 200, appeared 170 and qualified 119. The total registered is 750, appeared 665 and qualified 449. Absentees total 750−665=85. Those who appeared but did not qualify total 665−449=216. These two groups are disjoint; together they form the 301 registered candidates who did not qualify. A has the most qualified candidates in absolute number, 180. C has the highest qualification rate among appeared candidates: 119/170=70%, compared with A 180/270≈66.7% and B 150/225≈66.7%. Now suppose C’s qualified entry is hidden but the all-centre qualified total 449 is given. Subtract A and B: 449−180−150=119. This is a missing-data calculation, not an estimate. After solving, check 119≤170≤200. If the recovered qualified count exceeded C’s appeared count, either the total or the column had been misread.
Diagnostic worked case: ranked cumulative data
Five ranked categories have values 90, 70, 60, 50 and 30. The cumulative sequence is 90, 160, 220, 270, 300. The first three contribute 220/300≈73.3% of the whole. Categories three and four together contribute 270−160=110. The fourth category alone contributes 270−220=50. These differences show how a cumulative column encodes exact subsets: subtract the total before a group from the total at its end. The values are sorted high to low, but the cumulative numbers must rise. If a table labelled cumulative falls from one row to the next while all category values are non-negative, inspect the column alignment before calculating. To reach at least 70% of the whole, two categories are insufficient (160/300≈53.3%) and three are sufficient (220/300≈73.3%). If a question asks for the smallest number of top-ranked categories needed, test cumulative totals in order and stop at the first that crosses the threshold. This is different from a funnel, in which the same people pass through successive stages and counts usually fall.
Mixed-set self check before the exam
Take a ten-question sample containing at least one table, grouped bar chart, histogram, line graph, pie chart, radar chart, caselet, missing-data table and cumulative table. For each question, record the exact chart values you used and one equation. After marking, separate reading mistakes from calculation mistakes. A reading mistake includes the wrong row, wrong year, wrong legend colour, wrong interval or wrong population. A calculation mistake includes the wrong base, premature rounding or a basic arithmetic slip. Rework only the failed step first, then solve the entire question again without looking at the key. In a timed session, a rough bound can prevent a costly reread: a part cannot exceed its whole, a count cannot be negative, a five-point mean must lie between the smallest and largest values, and a cumulative total of non-negative parts cannot decrease. These checks are not substitutes for the exact key; they make it easier to notice when a plausible-looking option is impossible. Keep the solved equation beside each error until you can reconstruct it unaided.
Short mixed drill with answers
Use this synthetic dataset: five equal-width histogram intervals have frequencies 6, 9, 15, 12 and 8. The total is 50. Question one: How many observations are in the first two intervals together? Answer: 6+9=15. Question two: What is the modal interval position? Answer: the third interval because 15 is the largest frequency; this says nothing about an exact observed value. Question three: What percentage of observations fall in the last two intervals? Answer: (12+8)/50=40%. Now a table shows Period 1 values A=60 and B=90, and Period 2 values A=75 and B=105. Question four: What is the combined increase? Answer: Period 1 total 150, Period 2 total 180, increase 30. Question five: What is the percentage increase in the combined total? Answer: 30/150=20%. Question six: Which unit grew by more in absolute terms? Answer: neither; both rose by 15. Their percentage growth differs because A starts at 60 and B at 90. A grows 15/60=25%; B grows 15/90≈16.7%. These six small questions train interval reading, totals and changing bases without requiring a new diagram.
How to use a timed practice set
For each 50-question chapter, make an initial untimed pass until the chart-reading steps feel automatic. On the next pass, use a timer and write one short expression for each item. If a chart contains five values but the question needs only two, copy only those two; unnecessary transcription creates extra opportunities for error. When stuck, identify the exact missing step: locating a labelled value, choosing a denominator, or carrying out arithmetic. Move on if the missing step is not resolved quickly and return after the straightforward items. Review should take longer than scoring: group missed items by cause and practise the same cause across different chart types. For example, after a wrong denominator in a pie question, also revisit a caselet rate and a percent increase from a line graph. Their pictures differ, but each error comes from using the wrong whole. The goal is to carry a reliable method from one display to another. Because the data are synthetic, do not memorise any numeric answer as a general fact about Andhra Pradesh or police recruitment. A useful review sheet has four columns: question ID, value copied from the display, expression used, and correction. For a missed question about the total of two series, the correction might say “read teal at Unit C, then blue at Unit C; add those two labelled values.” That is more actionable than writing “silly mistake.” Repeat the missed item with the explanation hidden and check whether you can state the needed chart coordinates before doing arithmetic. If you cannot, revise the legend and axis lesson first. For a missed percentage question, put the base in words beside the denominator. For example, 30/150 means “30-unit increase over the original 150 units,” while 30/180 uses the new total and answers a different question. This simple verbal check often catches a base reversal immediately. On the final mixed pass, aim for consistent method rather than a perfect first-attempt score. When a chart has exact labels, copy them. When it has only a scale, state the scale step and avoid claiming more precision than the display permits. When a caselet gives a nested population, draw arrows from registered to appeared to qualified and place the count at each stage. After finishing the set, recompute any answer that looks impossible: a negative count, a part larger than its whole, an average outside its data range, or a cumulative total that falls. These are signals to inspect the reading and the equation again. Keep your final review focused on errors that recur, and use the current recruitment notice for any official syllabus or pattern claim.