What Are Mean and Median in G2 Geography Fieldwork? | Keep the Unusual Value Visible Before Choosing an Average

An average compresses several observations into one summary. That is useful when you need to compare groups or describe a central value, but it removes detail. The best choice depends on what the question asks and what the observations represent.

Imagine an invented enquiry about pedestrian counts at selected sites. If you want the mean count across those observed sites, every valid count contributes to the arithmetic total. If you want a middle observed count less influenced by an unusually busy site, the median may be helpful. Neither alone explains why the busy site differs. State the question first, then select a summary and retain enough information to show variation. This makes the calculation serve the enquiry rather than letting the availability of a familiar formula decide what the investigation means.

Counts, distances and temperatures are numerical quantities, while categories such as park, pavement and shopping street are labels. You can count how often labels occur, but adding arbitrary category codes does not automatically produce a meaningful geographical average.

In an invented land-use survey, assigning park = 1 and shop = 2 does not mean the mean code measures an amount of land use. The numbers identify categories rather than a common measurable scale. Ordinal response categories also need careful treatment, as explained in the linked Likert guide. Before calculating, ask what the numerical difference between two values means. This simple check prevents a spreadsheet from returning a neat average for data whose coding does not support the interpretation being claimed.

A pedestrian count over five minutes is not directly comparable with a count over fifteen minutes unless the difference in observation duration is addressed appropriately. Similarly, litter per quadrat depends on the defined quadrat area. The unit is part of the quantity being averaged.

Write a label such as ‘pedestrians counted in each standard ten-minute window at the selected sites’. That gives the mean a clear meaning. If durations differ, a rate may be useful, but its calculation and interpretation must match the question. Do not average raw counts from unequal periods and present the result as a typical standard-period count. A fair comparison begins before the arithmetic, with compatible definitions of what each observation includes. If the original records are incompatible, explain and address that limitation rather than hiding it inside one summary number.

For numerical observations x, the arithmetic mean is their sum divided by the number of values. Each included observation contributes to the total, so changing one value changes the mean unless another change offsets it. Repeated values still count as separate observations when they come from separate recorded units.

Use an invented count set of 8, 9, 10, 11 and 42. The sum is 80 and there are five observations, so the mean is 16. The mean need not equal any observed value; no site in this set has a count of 16. That does not make the calculation wrong. It means the mean summarises the total across the five units. Interpretation should mention the uneven pattern rather than implying every sampled site is close to 16. A mathematically valid average can still be an incomplete description.

The median depends on ordered position. For the invented values 8, 9, 10, 11 and 42, the middle is the third value, 10. If the same observations arrive in the order 42, 8, 11, 9, 10, the median remains 10 after sorting.

Do not select the middle entry in the collection sheet unless the sheet is already ordered by the measured variable. Fieldwork records are often arranged by site number or time, which is a different ordering. Preserve the original sequence for spatial or temporal interpretation, and create a separate ordered list for finding the median. This lets you use the same data for two purposes without destroying location information. The sorted list answers a distribution question; the original map or record answers where and when the observations occurred.

With four invented observations 8, 9, 10 and 11, the two middle values are 9 and 10. Their arithmetic mean is 9.5, so the median is 9.5. This is the numerical-data convention used in the exercises here.

A median of 9.5 pedestrians does not imply that half a person was counted. It is a summary lying between the two central counts. The same distinction applies to the mean: a summary can be fractional even if individual observations are whole-number counts. Retain suitable precision and explain what the number summarises. If the task concerns ordered categories rather than numerical counts, do not automatically average category codes as though they were measurements. The variable’s meaning determines how a middle position should be reported.

The mode is the most frequent value or category. It answers a different question from the mean and median. A numerical dataset can have one mode, several equally frequent modes or no single most frequent value under the convention used.

In an invented litter record of 0, 1, 1, 2 and 6, the mode is 1 because it occurs twice. The median is also 1, while the mean is 2. Agreement between mode and median does not make the mean incorrect; the summaries describe different features. For a land-use category record, the modal category may be meaningful when a mean code is not. Choose the measure that matches the variable and purpose. Listing all possible averages is less useful than explaining what one or two well-chosen summaries reveal about the enquiry.

The range is maximum minus minimum. It gives a simple measure of spread, though it depends only on the two extremes. Two datasets can have the same mean and very different distributions, so an average alone may conceal a major geographical contrast.

Imagine invented site counts A: 1, 2, 3, 4, 40 and B: 8, 9, 10, 11, 12. Both sum to 50 and have mean 10. A has range 39, while B has range 4. Their medians are 3 and 10. The summaries indicate a highly uneven A set compared with the more clustered B set. A fair comparison reports that difference instead of saying the groups are identical because their means match. A dot plot or table can preserve the pattern with very little extra complexity.

An unusually distant observation may be described as a possible outlier, depending on context and any stated identification rule. It may reflect a recording mistake, an instrument problem, a different condition or genuine variation. Being unusual does not, on its own, establish which explanation is correct.

In the invented pedestrian set, 42 is much larger than 8, 9, 10 and 11. Check the site record: was the period longer, was a group counted twice, or is the site near a busy entrance? Those possibilities lead to different responses. A confirmed error should be corrected through a documented basis; a valid busy-site count should remain part of the evidence. The useful habit is to investigate before deciding. Do not erase a value merely because its presence makes the mean less convenient for the story you expected.

Return to the collection sheet, timestamps, units and method notes. Look for a misplaced decimal, a copied site identifier, a different duration or an instrument warning. These checks can establish a specific reason for correction rather than a preference for a smoother graph.

Suppose an invented entry was copied as 42 when the original tally clearly records 12. That provides a basis to correct the transcription. If the original tally is also 42 and the method was followed, the number is not disproved by its distance from the other values. Preserve an audit note explaining any change. In school work, the explanation can be brief: original entry, verified correction and reason. Transparent handling of data makes the final conclusion more trustworthy and allows another student to reproduce the processing.

An observed zero means the defined measurement was taken and nothing of the counted item was recorded. A missing observation means the value is unknown or unavailable. Treating both as zero changes the meaning of the dataset.

In the invented litter values 0, 1, 1, 2 and 6, the zero belongs in the calculation. Their sum is 10 and mean is 2. If the first quadrat was never observed, it should not be assigned zero merely to complete the table. With only the four known values 1, 1, 2 and 6, the mean of observed values is 2.5, and the missing quadrat remains a limitation. Those are different records. Label missing values clearly so the reader can distinguish no litter from no evidence.

The denominator for an arithmetic mean is the number of included numerical observations, not the largest site number or the number originally planned. If one of ten planned measurements is missing, a mean of the nine observed values divides by nine.

Also distinguish repeat readings from independent locations. Taking five readings at one site creates five numerical values, but it does not give the same spatial coverage as five separate sites. A summary can be calculated correctly while its geographical interpretation remains weak. In the method and result, say whether the average summarises readings, sites, time windows or respondents. This small label tells readers what received equal weight. Without it, an apparent neighbourhood average might actually be dominated by repeated observations at one convenient location.

An invented site A has one reading of 10, while site B has four readings of 20. Averaging all five readings gives 18, but averaging the two site means equally gives 15. The difference comes from the unit receiving equal weight: individual readings in the first calculation, sites in the second. Decide which matches the enquiry before processing. If the aim is an equal-site comparison, unequal repeat counts should not silently give one site four times the influence. Keep the observation structure in the dataset rather than stripping away site labels.

A frequency table compresses repeated values. If one observed count occurs three times, it contributes three observations to the denominator and three times its value to the sum. The frequency itself is not another measured count to be mixed into the data.

Use an invented record: litter count 0 has frequency 2, count 1 has frequency 3, and count 4 has frequency 1. There are six observations. Their total is 0×2 + 1×3 + 4×1 = 7, giving mean 7/6, approximately 1.17 items per sampled quadrat. The ordered data are 0, 0, 1, 1, 1, 4, so the median is 1. Expanding the table once is a helpful check when learning the method. It reveals why dividing by three table rows would use the wrong denominator.

Two group means cannot generally be averaged equally to recover the mean across all observations if the groups contain different numbers. Reconstruct each group’s total from its mean and count, add those totals and divide by the combined count.

In an invented example, group A contains two observations with mean 10, so its total is 20. Group B contains eight with mean 20, so its total is 160. Together the total is 180 across ten observations, giving a combined mean of 18. The unweighted mean of the two means is 15, which answers an equal-group-weighting question instead. Neither calculation should be mislabelled. State whether the purpose is to give each group equal influence or each observation equal influence, then use the corresponding calculation.

A group median does not supply the group’s sum or full ordered distribution. Averaging two medians does not generally produce the median of all the underlying observations. To find a pooled median, you normally need enough information about the combined ordered data.

Consider invented groups A: 1, 100, 101 and B: 2, 3, 4. Their medians are 100 and 3. The average of those medians is 51.5. However, the combined ordered data are 1, 2, 3, 4, 100, 101, whose median is 3.5. This counterexample shows exactly why the shortcut fails. The two original medians tell us about their groups’ middle positions, not the pooled middle. Use this as a reasoning check whenever a resource offers only summary statistics and asks for a result that may require more detail.

The invented site counts are 8, 9, 10, 11 and 42. Mean = 80/5 = 16; median = 10; range = 42−8 = 34. Four counts cluster between 8 and 11, while one is much larger. The large value pulls the mean above the median.

A useful answer might say: the median describes the middle observed count, while the mean reflects the full total including the unusually busy site. Do not conclude that the median proves the large count is wrong. If the enquiry concerns resource demand across these sampled sites, the high count may be important. If it concerns a central observed site, the median may help. Choose the interpretation according to the question and keep the unusual site visible so the reader can understand why the summaries differ.

Suppose the original field sheet confirms that 42 was a transcription error and should be 12. The corrected dataset is 8, 9, 10, 11, 12. Its total is 50, mean is 10, median is 10 and range is 4. The correction has a documented basis.

Now consider simply deleting 42 without such evidence. The remaining four values total 38, so their mean is 9.5 and median is 9.5. That produces a different dataset and a different result. The numerical change does not justify the deletion. Explain why an observation was corrected, excluded or retained using the method and evidence. If showing a sensitivity comparison, label both versions and the reason for considering them. This lets readers see how influential the value is without pretending that an unverified removal is a factual correction.

Group A has invented counts 1, 2, 3, 4, 40 and group B has 8, 9, 10, 11, 12. Both means are 10. A’s median is 3 and range is 39; B’s median is 10 and range is 4. These statistics show why equal means do not establish similar patterns.

For a geographical interpretation, look at where the counts occur. A may contain one busy interchange among quieter sites, while B may have more evenly distributed movement. That explanation remains hypothetical until the locations are supplied. State the numerical pattern first and then use map evidence to assess a possible mechanism. Do not invent an interchange merely to make the answer sound geographical. A well-supported answer connects the distribution to actual site information while distinguishing observed variation from a suggested explanation.

Return to invented groups of two observations averaging 10 and eight averaging 20. The combined mean is 18 because (2×10 + 8×20)/(2+8) = 180/10. The larger group receives more weight in an observation-level mean because it contains more observations.

An independent check is to invent compatible raw data: A can be 8 and 12, while B can be eight values of 20. Their sum is still 180 across ten values. This shows why the result is not halfway between 10 and 20. If the groups were the same size, the simple mean of their means would match the pooled mean. The condition matters. When evaluating a published comparison, ask whether group sizes are supplied and whether the stated weighting matches the claim being made.

Use three invented matched observation periods. Site A temperatures are 26, 27 and 29°C; corresponding site B temperatures are 28, 29 and 30°C. The paired B-minus-A differences are 2, 2 and 1°C, with mean 5/3, approximately 1.67°C.

A’s mean is 82/3, approximately 27.33°C, and B’s is 29°C. Their difference is also approximately 1.67°C because the same three paired periods are included with equal weight. That agreement is a useful arithmetic check. It does not mean pairing is unimportant: pairing preserves which observations belong together and supports interpretation of time-specific differences. If one site’s observation is missing for a period, do not silently compare mismatched sets as though the original pairing remained intact. State the subset used and the consequence for the comparison.

Rounding intermediate values too early can alter a final result. Keep sufficient working precision and then report the final summary at a level appropriate to the measurements and task. A calculated decimal does not imply that an instrument observed to that precision.

For invented temperatures recorded to 0.1°C, an average may produce several decimal places, but reporting all of them can imply more observational precision than is justified. Follow the task’s instructions and identify approximate values where relevant. For counts, fractional averages are acceptable summaries even when individual observations are integers. Explain the unit rather than forcing every mean to a whole number. Clear rounding is part of communicating the evidence: it lets another reader reproduce the calculation without mistaking calculator output for a claim about measurement certainty.

A small table or dot plot can reveal clustering, extremes and gaps that one average conceals. For spatial data, a map can also show whether unusual observations cluster near a particular feature. Choose a display that preserves the aspect of variation relevant to the question.

Do not reorder a distance graph by the measured value merely to make a smooth curve. Sorting is useful for finding a median, but a spatial graph should retain actual position. Similarly, a bar chart of site counts should use labelled sites and a clear scale. An average line can help comparison if its meaning is stated. The display and summary should work together: one compresses the data, while the other helps readers see what was lost in that compression and whether the average answers the geographical enquiry.

A mean can be calculated exactly for an incomplete or biased sample. Correct arithmetic does not prove that the sample represents the whole neighbourhood. The sampling frame, selection process, timing and missing observations determine how far the result can reasonably be extended.

Suppose an invented team records only the busiest entrance. A precise mean of repeated counts there does not become an average for every entrance. Adding more repetitions at that same point improves detail about the sampled setting but leaves spatial coverage limited. A useful evaluation names the missing variation and proposes a suitable extension. This connects statistical processing to the earlier fieldwork decisions. The quality of the conclusion depends on both: how observations were selected and how they were summarised after collection.

A larger sample can reduce some random variation while leaving systematic omissions unresolved. For instance, counting the same accessible entrance on twenty afternoons still excludes an entrance that is never observed. Describe this as a frame or coverage issue and propose observations that reach the missing part. The aim is to improve the evidence’s match to the intended population, rather than treating sample size as a universal repair for every weakness.

A compact conclusion can name the main pattern, cite a suitable average, identify important variation and limit the claim to the sampled units and period. For the invented five-site dataset, say that most counts lie between 8 and 11, with one count of 42 raising the mean to 16 above the median of 10.

Then use site evidence to explain the unusual observation or identify what additional information is needed. Avoid saying the whole district has a typical count of 16 without context. If evaluating methods, explain whether another summary, a fuller display or improved sampling would address the specific weakness. Those are different improvements. A transparent conclusion is useful even when the dataset is small, because it tells readers exactly what was found and what remains unresolved rather than concealing uncertainty behind a single impressive-looking average.

Is the median always better? It depends on the question and distribution. Should an outlier be removed? Investigate its origin and follow a defensible, documented rule. Does an observed zero count in the mean? Yes; a genuine zero is an observation, while a missing value is unknown. Can group medians be averaged to obtain a pooled median? Not in general.

For revision, calculate mean, median and range for the two equal-mean groups. Next explain why the busy site may be valid. Finally work through the unequal-group example and identify the weighting. These tasks develop more than formula recall: they connect calculation to the meaning of the observations. A parent can help by asking what each number represents and what the average leaves out. The student’s explanation should be clear enough that someone else could reconstruct the dataset and understand the conclusion.