An animal can spend most of its recorded time in a habitat and still use that habitat less than its availability would suggest. Conversely, a habitat with fewer recorded visits can be used disproportionately often when it occupies only a small part of the animal’s defined available environment.
The short answer is to compare use with an explicitly defined availability measure, at the same scale and for the same animal or population question. Then keep that pairing intact when combining records. Pooling many observations from one animal against an equally weighted map of several animals can create an apparent preference that none of the individual comparisons supports.
This guide develops original, invented two-habitat examples. It teaches arithmetic and interpretation, not a method for tagging animals, entering habitats or deciding conservation priorities from a small worksheet. The supplied use proportions are stipulated observations under a stated sampling scheme; they are not automatically continuous residence times or independent biological replicates.
The Wildlife subject guide connects responsible observation, habitats and conservation evidence. Here we ask a narrower question: what must be matched before a habitat-use percentage becomes a meaningful comparison?
eduKateSG · Wildlife
Find your next learning step
Choose the question closest to your learning need, or follow the worked examples in order. Each route connects a clearly stated model with a changed-context check.
ROUTE 2 · CHAPTERS 5–8
Test ratios and pooling
Change the reference and expose mismatched pooled weights.
ROUTE 3 · CHAPTERS 9–12
Choose the target and weights
Compare pooled records, equal-animal summaries and a small model.
ROUTE 4 · CHAPTERS 13–16
Preserve uncertainty and scope
Handle classification, detection and ecological interpretation.
ROUTE 5 · CHAPTERS 17–20
Practise and continue
Solve a new example, repair errors and preserve the evidence card.
Chapter contents
Define a paired comparison · Chapters 1–4
1. Separate recorded use from the comparison you want to make
2. Define what available means before drawing the boundary
Test ratios and pooling · Chapters 5–8
5. Change availability while keeping every observation fixed
6. Compare relative ratios without calling them probabilities
7. Introduce two animals with different available environments
Choose the target and weights · Chapters 9–12
9. Match weights when the target is the pooled record
10. Give each animal equal weight when that is the question
11. Use a small model to distinguish shares from selection weights
Preserve uncertainty and scope · Chapters 13–16
13. Preserve classification uncertainty at habitat boundaries
14. Keep detection and dependence outside the ratio's promises
Practise and continue · Chapters 17–20
17. Diagnose a convincing but mismatched explanation
18. Build an evidence card that preserves the pairing
19. Questions about majority use and habitat preference
20. Follow the question into movement and conservation evidence
CHAPTER 1 OF 20 · Define a paired comparison
1. Separate recorded use from the comparison you want to make
Suppose sixty of one hundred accepted location records fall in wet habitat and forty in dry habitat. The first defensible calculation is a recorded-use share: sixty per cent of those records were classified wet. The word preference adds a comparison that the count alone does not contain.
If wet habitat covers almost everything accessible under the study definition, sixty per cent could be relatively low. If wet habitat is scarce within that definition, the same sixty records could be relatively high. The numerator has not changed; the relevant context has.
Begin with three plain statements: which records count as use, which alternatives are treated as available, and which entity the comparison represents. An individual during one period, a collection of tracked animals and a species across a region are different targets. Giving them the same label does not make their denominators interchangeable.
This separation also prevents an emotional reading of the terms. A lower-than-availability use share does not mean an animal dislikes a habitat in a human sense. It describes a relationship between a supplied use measure and a supplied comparison distribution. Behavioural mechanisms, constraints and benefits require further evidence.
A learner should be able to state the sixty-per-cent observation without mentioning availability, then explain what extra information is needed before making the comparison. That small distinction is the foundation for every calculation that follows.
CHAPTER 2 OF 20 · Define a paired comparison
2. Define what available means before drawing the boundary
For our first exercise, availability is the fraction of a stipulated accessible study area assigned to each of two exhaustive habitat classes. The classes do not overlap, and their availability fractions sum to one. This is a supplied teaching definition, not a claim that area is always the correct measure.
A whole island, an animal’s seasonal range and the places reachable between two fixes provide different alternatives. A polygon on a map may include barriers or places the animal could not reach during the relevant period. Calling every mapped cell available can therefore change the question substantially.
The primary research record Practical Guidance on Characterizing Availability reports that the spatial extent and sampling of availability affect habitat-selection inference. Our examples below isolate that dependence with invented arithmetic; they do not reproduce the study’s animal data or fit its models.
Write the boundary rule in a way another reader could apply. Include the period and habitat classification, and state whether accessibility was supplied, modelled or unknown. A precise percentage calculated from an unexplained boundary still leaves the most consequential assumption hidden.
For classroom learning, use fictional maps or already approved non-sensitive material. The purpose is to examine a comparison, not to obtain more intrusive data. Analytical curiosity does not create permission to disturb wildlife or disclose protected locations.
CHAPTER 3 OF 20 · Define a paired comparison
3. Keep a location share distinct from a time share
A record at one moment establishes a classified location under the observation’s accuracy assumptions. It does not directly establish how long the animal remained there between records. Regularly spaced accepted observations can support a particular sampling interpretation, but gaps and location-dependent failures complicate that interpretation.
Imagine ten accepted records in wet habitat and ten in dry habitat. If the wet records are separated by ten minutes while the dry records are separated by an hour, simply counting twenty fixes does not allocate equal observed time to the two habitats. The record count and duration question need different accounting.
The existing guide on habitat residence time between tracking fixes develops that separate issue. This lesson begins after a use measure has been chosen and supplied. We do not convert our hypothetical fix counts into exact time budgets.
That boundary keeps the new comparison honest. A use-to-availability calculation cannot repair biased or misunderstood use data. First establish what the numerator represents; only then decide which availability distribution is relevant to it.
Ask the learner to attach a unit-like label to each share: accepted-fix share, stipulated time share or another defined measure. Percentages can hide these distinctions because they all look dimensionless. Their interpretation still depends on the process that produced the numerator and denominator.
CHAPTER 4 OF 20 · Define a paired comparison
4. Calculate the first two-habitat example
Return to one hundred accepted records, sixty wet and forty dry. Supply availability of eighty per cent wet and twenty per cent dry for the same animal and period. Define a simple descriptive ratio w as use share divided by availability share.
For wet habitat, w is 0.60/0.80=0.75. For dry habitat, it is 0.40/0.20=2. The wet habitat contains most recorded use, yet its share is below its supplied availability. Dry habitat contains fewer records, yet its use share is twice its availability share.
| Habitat | Accepted records | Use share | Supplied availability share | Use divided by availability |
|---|---|---|---|---|
| Wet | 60 | 0.60 | 0.80 | 0.75 |
| Dry | 40 | 0.40 | 0.20 | 2.00 |
The table does not rank habitat quality or prescribe habitat removal. It describes the selected comparison. An animal might obtain an essential resource during a short visit, face constraints elsewhere, or respond to conditions not represented by our two labels. None of those mechanisms is identified by division alone.
Check the arithmetic through expected counts under proportional use: eighty wet and twenty dry in one hundred records. The supplied record has twenty fewer wet and twenty more dry than that benchmark. This comparison is descriptive; it is not a significance test, because we have not supplied an appropriate sampling model or independence assumptions for a test.
CHAPTER 5 OF 20 · Test ratios and pooling
5. Change availability while keeping every observation fixed
Keep the sixty wet and forty dry records. Change only the stipulated available area to twenty per cent wet and eighty per cent dry. Wet’s ratio becomes 0.60/0.20=3, while dry’s becomes 0.40/0.80=0.50.
The recorded use has not changed at all. The comparison has reversed because the reference distribution changed. This is why an unexplained change to a study boundary can alter the conclusion without adding a single wildlife observation.
Neither boundary is validated by its arithmetic. If one includes unreachable terrain while the other follows a justified accessible region, their scientific value differs. If both represent legitimate but different questions, each result should be labelled with its own scope rather than treated as a disagreement about one universal preference number.
Try a third supplied wet availability of sixty per cent. With wet use at sixty per cent, both habitat ratios become one. That is proportional use for this defined two-class comparison. It does not prove random movement, indifference, or equal survival benefits. Many processes could produce the same aggregate shares.
The transferable skill is conditional reasoning. State the conclusion as “relative to this availability definition, under this use measure.” That wording is not an excuse for choosing arbitrary boundaries. It makes the assumption visible so another reader can challenge whether it answers the ecological question.
CHAPTER 6 OF 20 · Test ratios and pooling
6. Compare relative ratios without calling them probabilities
In the first example, wet’s ratio 0.75 divided by dry’s ratio 2 gives 0.375. This relative comparison summarises the wet-to-dry use odds in relation to the wet-to-dry availability odds. It is not a probability that the animal chooses wet habitat at its next decision.
To check, recorded wet-to-dry odds are 0.60/0.40=1.5. Available wet-to-dry odds are 0.80/0.20=4. Dividing 1.5 by 4 again gives 0.375. The two routes agree because they rearrange the same supplied fractions.
A ratio above one is permitted. It describes disproportionate use, not a probability exceeding one. Likewise, the two simple w values need not sum to one:0.75+2=2.75 in our first table. They have different availability denominators.
If a report normalises such ratios to sum to one, that is an additional transformation. It should be named and interpreted as such, not silently exchanged with the original use shares. A normalised weight is not automatically the observed fraction of records in a habitat.
For practice, take the changed-availability case with ratios 3 and 0.5. Their relative ratio is 6, whereas their sum is 3.5. Normalising gives 6/7 and 1/7. Neither value equals the original observed wet share 0.60. Keeping these quantities separate prevents a clean-looking chart from replacing one question with another.
CHAPTER 7 OF 20 · Test ratios and pooling
7. Introduce two animals with different available environments
Now consider Animal A and Animal B. Each has a separately supplied available environment. For A, wet availability is 0.80 and recorded wet-use share is 0.70. For B, wet availability is 0.20 and wet-use share is 0.10.
A’s wet ratio is 0.70/0.80=0.875. B’s is 0.10/0.20=0.50. Each individual comparison shows wet use below its own supplied availability. The dry ratios are 0.30/0.20=1.50 for A and 0.90/0.80=1.125 for B.
These are not two observations of one identical environment. A lives within an availability distribution dominated by wet habitat; B’s supplied environment is dominated by dry habitat. Combining them requires deciding what the combined question should represent.
Suppose the intended target is the average recorded-use share of these two individuals, giving each individual equal weight. That is a legitimate descriptive question. Suppose instead the target is the habitat distribution of all accepted fixes in the dataset. That is another legitimate question. They agree only under particular conditions.
Before revealing record counts, ask the learner to predict whether pooling can preserve every individual comparison automatically. The answer is no: weights matter. The next chapter supplies deliberately unequal counts so the consequence can be inspected rather than hidden behind the word average.
Let Animal A contribute 900 accepted fixes and B contribute 100. Their wet records are 630 and 10 respectively, giving 640 wet records out of 1000. Pooled recorded wet use is therefore 0.64.
Now make the deliberate error of comparing that pooled use with the simple average of the two availability shares: (0.80+0.20)/2=0.50. The resulting ratio is 0.64/0.50=1.28, apparently above proportional use. Yet both individual wet ratios were below one.
| Animal | Accepted fixes | Wet-use share | Wet-availability share | Individual wet ratio |
|---|---|---|---|---|
| A | 900 | 0.70 | 0.80 | 0.875 |
| B | 100 | 0.10 | 0.20 | 0.500 |
| Pooled-use shortcut | 1000 | 0.64 | Unmatched mean 0.50 | Misleading 1.28 |
The reversal arises because use was weighted nine-to-one by fix count while availability was weighted one-to-one by animal. The two sides describe different mixtures. It is not evidence that an individual changed behaviour when the rows were added.
This example is deliberately constructed to reveal the weighting error. It does not demonstrate that every pooled wildlife analysis is invalid or that all animals should always receive equal weight. The right response is to state the target and align the comparison with it, not to replace one universal pooling rule with another.
CHAPTER 9 OF 20 · Choose the target and weights
9. Match weights when the target is the pooled record
If the target is all 1000 accepted fixes, a matching availability mixture weights A’s supplied environment by 900/1000 and B’s by 100/1000. Wet availability becomes 0.9×0.80+0.1×0.20=0.74.
The paired descriptive wet ratio is now 0.64/0.74=32/37, approximately 0.865. Dry use is 0.36 and matched dry availability 0.26, giving 18/13, approximately 1.385. The result no longer creates the earlier false above-availability wet comparison.
These matching weights answer the stipulated pooled-record question. They do not make a thousand fixes equivalent to a thousand independent animals. If the research target concerns variation among animals, the dependence structure and sampling design still need an appropriate analysis.
Explain why the matching works here. Every fix assigned to A carries the same supplied available fraction 0.80 for this simplified period. Weighting those fractions by the number of contributing fixes preserves the mixture that produced the pooled use share. If availability changes across an individual’s observations, a single animal-level fraction may no longer represent the correct pairing.
That last condition points to the next useful research question. Matching a denominator is necessary for this descriptive exercise, but it does not automatically validate a complex availability model. The learner should preserve the scope of the repair rather than announce that matching weights resolves every ecological limitation.
CHAPTER 10 OF 20 · Choose the target and weights
10. Give each animal equal weight when that is the question
For the alternative equal-animal target, average A’s wet-use share 0.70 and B’s 0.10 to obtain 0.40. Average their supplied wet availability 0.80 and 0.20 to obtain 0.50. The ratio of these two equally weighted means is 0.80.
That is different from both the misleading 1.28 and the matched pooled-record ratio of approximately 0.865. It is also different from the average of the two individual wet ratios: (0.875+0.50)/2=0.6875.
All three correctly computed quantities can appear in the same worksheet: a ratio of equally weighted means, a ratio under fix-count weights, and a mean of individual ratios. They have different mathematical structures. A report should not switch among them simply because one produces a more dramatic result.
The mean of ratios gives each individual’s ratio equal weight. The ratio of means pools numerator and denominator summaries before dividing. Division does not generally commute with averaging, particularly when denominators differ. Writing out the two individual denominators makes the distinction easier to see.
Ask the student to name the target before calculating. If they cannot explain whether the result represents records or individuals, supply no more decimal places. The missing ingredient is a question definition, not numerical precision. Once the target is stated, the calculation can be selected and its limitations discussed.
CHAPTER 11 OF 20 · Choose the target and weights
11. Use a small model to distinguish shares from selection weights
An invented weighted-choice model makes another distinction visible. Suppose wet habitat has a fixed relative weight 3 and dry habitat weight 1. If wet availability is a, the model allocates wet use as 3 a divided by[3 a+(1−a)]. This is a stipulated mathematical model, not a fitted description of a real animal.
With a=0.20, predicted wet use is 0.60/(0.60+0.80)=3/7, about 42.9 per cent. With a=0.80, it is 2.40/(2.40+0.20)=12/13, about 92.3 per cent. The relative weights stayed the same while the predicted use shares changed substantially.
Even the simple wet use-to-availability ratio changes: (3/7)/0.20=15/7 in the first environment, and(12/13)/0.80=15/13 in the second. Those changing ratios do not imply that we changed the model’s relative weight of 3.
The relative wet-to-dry ratio remains 3 in both constructed cases. Check it using the observed-odds-to-available-odds calculation from chapter 6. This gives a concrete reason to avoid comparing raw use shares, simple ratios and model weights as if they were interchangeable descriptions.
The model also shows why observed use is not an isolated property of the animal. Under its own assumptions, use results from weights interacting with availability. Real habitat-selection analysis needs richer decisions about the processes, data and scale involved.
CHAPTER 12 OF 20 · Choose the target and weights
12. Handle zero and very small availability honestly
Division by zero is not repaired by printing a large number. If wet availability is stipulated as zero while a verified wet-use record exists in the same comparison, the record and availability definition conflict. Investigate mapping, classification, boundaries, accessibility assumptions and timing.
Perhaps the wet record lies outside the chosen availability polygon, or habitat changed during the period. Perhaps location uncertainty crossed a boundary. The ratio cannot diagnose which explanation is correct. Preserve the discrepancy rather than deleting the record to make the arithmetic work.
If availability is positive but small, the ratio can be large and sensitive to small denominator changes. A use share 0.02 gives ratio 2 when availability is 0.01, but ratio 1 when availability is 0.02. That comparison does not establish whether either availability estimate is credible.
For a classroom sensitivity exercise, supply an availability range 0.01 to 0.02 rather than a single exact value. With use fixed 0.02, the ratio lies between 1 and 2. The direction of the bound follows because dividing by a larger positive denominator reduces the result.
A zero observed use share with positive availability yields a descriptive zero ratio, but it does not prove ecological impossibility. Sampling and detection still matter. These cases should remain separate: a mathematical denominator failure, a sensitive positive denominator, and an observed numerator of zero carry different information.
CHAPTER 13 OF 20 · Preserve uncertainty and scope
13. Preserve classification uncertainty at habitat boundaries
Suppose 100 accepted fixes include fifty definitely wet, thirty definitely dry and twenty unresolved at a boundary. Without additional assumptions, the wet recorded-use share lies between 0.50 and 0.70. Reporting 0.60 would require a rule allocating half the unresolved records to wet.
If availability is stipulated at 0.80 wet, the simple wet ratio lies between 0.625 and 0.875. The whole range is below one under that supplied denominator. If availability were 0.60 instead, the ratio range would be 5/6 to 7/6, crossing one. The classification uncertainty would then affect even the directional description.
These are deterministic bounds from the stipulated unresolved records. They are not confidence intervals and do not include every uncertainty in the map, sample or ecological process. Label them as bounds under the supplied classification information.
A useful evidence card distinguishes uncertainty in use classification from uncertainty in availability. One concerns which habitat the recorded location belongs to; the other concerns the alternatives represented by the denominator. Combining them without labels makes the result harder to repair.
The learner’s task is to identify which ambiguity matters for the stated conclusion. A stable below-one bound under one supplied availability may support a narrow descriptive comparison. It cannot validate the boundary or prove the animal’s motivation. The result remains conditional on the whole measurement contract.
CHAPTER 14 OF 20 · Preserve uncertainty and scope
14. Keep detection and dependence outside the ratio's promises
An animal’s presence can go unrecorded. A device or observation process can also be less successful in one habitat than another. If wet locations are more often missing, the accepted-fix share can differ from the underlying location process even with flawless arithmetic.
The existing guide on detection probability, repeat surveys and false absence owns the broader presence-versus-detection question. Here the practical connection is to check whether the chosen use records represent the phenomenon the habitat comparison is meant to describe.
Repeated fixes from one animal can also be strongly related. A sequence of locations near one resting place does not create a new independent animal with each timestamp. Increasing the recording frequency can increase the weight of that animal in a naïve pooled analysis without increasing the number of biological individuals.
Our examples intentionally avoid a confidence test or a claim of statistical significance. A competent analysis would need the sampling and dependence structure, not just the final ratio. The row counts alone do not supply that model.
For a learning check, change A’s record count from 900 to 1800 while preserving its use proportion. Predict which summaries change. The individual ratios and equal-animal summaries stay fixed. A pooled-record summary shifts toward A. That is an arithmetic consequence of the target’s weights, not evidence of newly observed animals.
For that changed-count exercise, A now contributes 1260 wet records and B contributes 10. Pooled wet use becomes 1270/1900, approximately 0.668. The corresponding fix-weighted wet availability is(1800×0.80+100×0.20)/1900=1460/1900, approximately 0.768. Their ratio is 127/146, approximately 0.870. All individual shares and ratios remain as supplied.
The revised pooled value is closer to A’s individual comparison because A supplies an even larger fraction of the accepted records. Nothing in this arithmetic establishes that the additional fixes contain independent ecological information. That question would require the actual spacing, observation process and analysis model. The example shows exactly which descriptive summary is sensitive to extra records from an existing individual, giving a student a concrete way to challenge an unexplained change in the word average.
CHAPTER 15 OF 20 · Preserve uncertainty and scope
15. Keep use, benefit and conservation value distinct
A place used frequently is not automatically the place contributing most to survival or reproduction. A place used briefly can provide an essential resource. Our percentages do not contain demographic outcomes, hazards or the full sequence of behaviours.
The primary paper The Interpretation of Habitat Preference Metrics explains that inference depends on use, availability and modelling choices; a preference metric is not a direct reading of intrinsic motivation or habitat value. Our worksheet supplies no evidence that would overcome those limits.
Imagine a classroom case in which a dry patch contains a short drinking visit while wet habitat supplies many resting records. The story is an illustration of two possible functions, not an explanation of our earlier invented dataset. We must not invent behaviour after seeing the ratios and then present it as a discovered cause.
To evaluate benefit, ask for evidence tied to the actual conservation objective. Persistence, reproduction, safe movement and access to essential resources require different observations. An appealing label such as preferred cannot replace that chain of reasoning.
The Wildlife Conservation guide provides a broader route into threats, outcomes and responsible intervention. Follow that route when the question moves beyond descriptive habitat comparison. A numerical use ratio can inform a question while remaining insufficient to settle it.
CHAPTER 16 OF 20 · Preserve uncertainty and scope
16. Reconstruct a changed habitat comparison independently
Try a new animal with 200 accepted records: thirty in wooded habitat and 170 in open habitat. Supply available fractions 0.10 wooded and 0.90 open for the same period and boundary. Name the measure before dividing.
The wooded use share is 0.15 and its simple ratio is 1.5. Open use is 0.85 and its ratio is 17/18, about 0.944. Most records are open, while wooded use is above its smaller availability share. The relative wooded-to-open ratio is 27/17, about 1.588.
Now change wooded availability to 0.30 and open to 0.70, keeping the records fixed. Wooded’s ratio becomes 0.50; open’s becomes 17/14, about 1.214. Explain the reversal without claiming that any animal moved. The reference distribution changed.
For a third task, keep the original 0.10 wooded availability but mark ten of the 200 records unresolved rather than definitely open. With thirty definitely wooded and 160 definitely open, wooded use lies between 0.15 and 0.20. Its ratio lies between 1.5 and 2. This classification bound does not incorporate uncertainty in the supplied availability.
A complete answer states the target, paired availability, arithmetic and limited conclusion. Once those elements are secure, ask the student to invent a different plausible weighting question and explain what extra data it would require. Transfer means recognising the structure while changing the context, not copying the wet-and-dry labels.
CHAPTER 17 OF 20 · Practise and continue
17. Diagnose a convincing but mismatched explanation
Consider this claim: “Wet habitat received sixty-four per cent of all records and occupies half the average map, so the animals selected it.” The numbers refer to our unequal 900-and 100 example. The first unstable step is pairing fix-weighted use with equally weighted availability.
Repair that step before arguing about terminology. Under the pooled-record target, use matched wet availability 0.74. Under the equal-animal target, use equally weighted wet use 0.40. Neither repair supports the shortcut’s ratio 1.28.
A second student writes that A’s wet ratio 0.875 is an 87.5-per-cent probability of entering wet habitat. The computed ratio is correct, but the probability interpretation is unsupported. Return to the definition: use share divided by availability share.
A third student computes all ratios correctly and recommends removing low-ratio habitat. That conclusion adds an intervention and a welfare claim absent from the supplied evidence. The repair is to separate descriptive comparison from ecological benefit and decision-making.
This approach to feedback preserves correct work while correcting the actual reasoning break. Ask the learner to explain their revised result aloud, then solve the wooded-habitat task without prompts. A polished paragraph matters less than knowing which denominator belongs to which numerator and which claim the completed calculation can support.
CHAPTER 18 OF 20 · Practise and continue
18. Build an evidence card that preserves the pairing
A reproducible record names the animal or population target, observation period, use measure, habitat classes and classification rule. It then states the availability definition, spatial boundary, accessibility assumptions and whether that denominator changes through time.
Record accepted and missing observations separately. Preserve uncertain classifications rather than silently forcing them into the preferred category. Keep original locations protected according to the project rules; a public learning example usually does not need precise real coordinates.
For aggregation, write the weights explicitly on both sides. Are records, animals, days or another unit being averaged? Is the final quantity a ratio of means, a mean of ratios or a fitted model parameter? A one-line formula beside the label can prevent substantial misunderstanding.
Finally, include the sensitivity checks that affect interpretation: alternative justified availability boundaries, unresolved habitat classifications, missingness and the influence of uneven record counts. Do not call this list a complete field protocol or statistical analysis plan. It identifies the information needed to understand the particular descriptive comparison.
The strongest student explanation is reconstructible without access to the student’s intention. Another reader should be able to recover every supplied share, weight and transformation, then see exactly where the explanation stops. That is a practical standard of scientific communication even when the dataset is fictional and the calculation remains elementary.
CHAPTER 19 OF 20 · Practise and continue
19. Questions about majority use and habitat preference
Can the most-used habitat have a ratio below one?
Yes. In our first example, wet use is 0.60 against availability 0.80, giving 0.75. Majority use and above-availability use answer different questions.
Should every animal always receive equal weight?
No universal weighting rule answers every target. State whether the question concerns an average individual, the pooled records or another population quantity, then preserve a matching comparison and appropriate analysis.
Does matching weights solve sampling bias?
It repairs the specified mismatch in our descriptive exercise. It does not establish unbiased use data, accessible habitat boundaries, independent observations or a valid confidence analysis.
Can a ratio reveal why the animal used a habitat?
Not by itself. The comparison does not identify motivation, benefit or mechanism. Different processes can produce the same shares.
What if availability is zero but use is observed?
The ratio is undefined. Investigate the joint record and availability definition rather than substitute an arbitrary tiny denominator or delete the observation.
CHAPTER 20 OF 20 · Practise and continue
20. Follow the question into movement and conservation evidence
For the wider purpose and ethical responsibilities, use the Wildlife subject guide and Wildlife Conservation. They connect careful observations with ecological questions and bounded conservation claims.
For distinct measurement problems, follow habitat residence time between fixes and repeat surveys with imperfect detection. One asks about time between locations; the other separates recorded detection from site state. Neither is replaced by a use-to-availability ratio.
For primary research context, read the accessible records for Northrup and colleagues on availability, Johnson's comparison of usage and availability and Beyer and colleagues on interpretation. The links support the conditional nature of the comparison; all worksheet numbers and pooling counterexamples here are original teaching constructions.
Return through The Gold Standard of Education and How X Works. Carry forward the habit that makes these examples useful: keep each observation attached to the alternatives, weights and question that give it meaning.