THC Subject Library
Evidence, Measurement & Research
Learn to separate observation from conclusion, choose measurements that answer the actual question, understand replication and sampling, audit claims, preserve uncertainty, and build cultivation records that can be compared.
Guided study · Foundation
What do we actually know, what was measured, and what conclusion is justified by that evidence?
Use this question to organize the literature below. The goal is to connect observation to plant function before jumping to a correction.
Measure first
Evidence to collect
- What instrument, sampling method, location, timing, units, and calibration state produced the measurement?
- How much biological and measurement variation exists across repeated observations?
- Which variables changed together and which were held reasonably constant?
Interpret carefully
Common reasoning errors
- Treating a single reading as a stable condition.
- Confusing correlation, sequence, or improvement after an intervention with proof of cause.
- Reporting more precision than the sensor, sampling method, or experiment can support.
Apply it
Test one cultivation assumption
- Write one clear question and the observation that would count as evidence.
- Choose one variable to change and identify the major conditions you will try to keep comparable.
- Collect repeated measurements using the same method and units.
- Summarize what the data support, what they do not prove, and what next observation would reduce uncertainty.
Encyclopedia depth
Go deeper after the subject overview.
This subject page teaches the model. The encyclopedia hubs break that model into narrower reference lessons.
Core literature
Build the model before making the decision.
The sections below keep plant science, observation, and practical checkpoints together so the page works as a usable reference instead of a text dump.
How to study Evidence, Measurement & Research
Reliable cultivation knowledge depends on defined questions, repeatable methods, representative sampling, records, comparison, and uncertainty. Better measurement reduces confident but unsupported conclusions.
Common interpretation trap: Confusing precision with accuracy, correlation with causation, or one successful run with a general rule.
- Question: What exact claim or question is being tested?
- Question: How was the observation or measurement collected?
- Question: What alternative explanations, missing data, or sources of bias remain?
- Record: method and instrument identity
- Record: sampling location and timing
- Record: raw observations
- Record: replicates or repeated observations
- Record: changes and outcomes
Evidence is claim-specific
Evidence is strong only in relation to a particular claim. A photograph can show visible appearance, but it may not identify a hidden cause. A laboratory test can quantify a sample, but it may not represent the whole crop. A controlled experiment can estimate an effect under its tested conditions, but extrapolation beyond those conditions introduces uncertainty.
The first step in evaluating a claim is therefore to state exactly what is being claimed and what evidence would distinguish it from reasonable alternatives.
- Write the claim in one testable sentence.
- List plausible alternative explanations.
- Match evidence type to the claim rather than to convenience.
Observation, measurement, and inference
Observation describes what was seen or recorded. Measurement assigns values using a defined method. Inference is the interpretation drawn from those observations and measurements. Keeping these layers separate prevents confident conclusions from becoming disguised as raw facts.
For cultivation records, preserve original values and photographs before summarizing them. A later analyst should be able to see what was actually measured.
- Label observations separately from interpretations.
- Preserve raw values and units.
- Record instrument, method, time, and location.
Sampling, replication, and pseudoreplication
A sample should represent the population or condition about which a claim is made. Multiple measurements from the same plant or container can improve precision for that unit but are not necessarily independent biological replicates.
Replication captures independent experimental units. Subsampling captures variation within an experimental unit. Confusing the two can make evidence appear stronger than it is.
- Identify the true experimental unit.
- Distinguish biological replicates from repeated readings.
- Describe how sample locations were selected.
Measurement quality and uncertainty
Every measurement has limitations. Calibration, resolution, sensor placement, sample handling, drift, environmental interference, and operator technique can affect results. More decimal places do not automatically mean more accuracy.
Useful records include units, method, calibration state, and uncertainty when known. When uncertainty is not quantified, it should still be acknowledged in interpretation.
- Use appropriate calibration and quality-control checks.
- Do not report precision the method cannot support.
- Repeat suspicious measurements before building a major decision on them.
Claim auditing and monitored decisions
A claim audit asks who made the claim, what exactly was tested, what controls and comparisons were used, whether the result was replicated, whether the statistics answer the question, and how closely the evidence matches the intended cultivation context.
Cultivation decisions can use the same discipline. Record baseline conditions, make a defined change, specify what response would count as improvement or harm, and monitor the result. This turns everyday management into a learning system without pretending that every grow is a formal experiment.
- Record the baseline before changing the system.
- Define expected response and observation window.
- Keep negative and null results instead of only memorable successes.
Controls, randomization, and blocking
A comparison becomes easier to interpret when it includes an appropriate control or reference condition. The control establishes what happened without the tested change under the same general period and environment. Without it, a before-and-after difference can be confused with normal plant development, weather, equipment drift, or another simultaneous change.
Randomization reduces systematic placement bias by preventing the researcher from intentionally or unintentionally assigning one treatment to the most favorable positions. Blocking can improve comparisons when a known gradient exists, such as distance from a light source, irrigation zone, bench position, or outdoor slope. Treatments are then compared within blocks that share similar background conditions.
- Define the control or reference condition before the trial starts.
- Randomize treatment assignment when the design allows it.
- Use blocks when a known spatial or environmental gradient could confound the treatment.
- Record the assignment method so the design can be audited later.
Effect size, variability, and practical importance
A difference between treatments has at least two separate questions: how large is the observed effect, and how uncertain is that estimate? Reporting only whether a statistical test crosses a threshold can hide the size and variability of the response. Raw measurements, treatment summaries, spread, sample size, and an interval estimate provide more information about what the data can support.
Practical importance is also contextual. A small average change may matter if it is repeatable, inexpensive, and linked to an important outcome, while a larger but highly variable change may be unreliable. The decision should connect magnitude, uncertainty, biological meaning, cost, risk, and repeatability rather than reducing evidence to one p-value.
- Report the direction and magnitude of the observed difference.
- Keep variability and sample size beside treatment averages.
- Separate statistical evidence from biological or operational importance.
- Preserve null and negative findings so future decisions are not biased toward memorable successes.
Time series, baselines, and change attribution
Plant systems change through time even when management is stable. A useful baseline therefore contains enough observations to show the normal direction and variability before an intervention. One measurement immediately before and one immediately after a change can miss delayed responses, daily cycles, sensor noise, acclimation, and developmental trends.
Time-series records are strongest when measurement timing and methods remain consistent and when important external events are timestamped. Irrigation, pruning, transplanting, equipment changes, environmental excursions, pest treatments, and developmental transitions can all alter the trajectory and should be visible on the same timeline.
- Collect a meaningful baseline before changing the system when possible.
- Keep sampling time and method consistent across the comparison.
- Timestamp interventions and environmental excursions.
- Look for sustained trajectory changes rather than reacting to one noisy point.
Reproducibility, metadata, and data integrity
A result is more useful when another person can understand how it was produced. Metadata should preserve plant or batch identity, location, date and time, developmental stage, treatment definition, instrument, calibration state, units, sampling method, environmental context, and any deviations from the planned method. This information often determines whether two datasets can legitimately be compared.
Data integrity means preserving raw observations, corrections, exclusions, and transformations transparently. Do not silently overwrite a surprising value because it looks wrong. Flag it, investigate the instrument or method, repeat the measurement when appropriate, and retain the original record with the reason for any exclusion or correction.
- Use stable IDs for plants, treatments, samples, and instruments.
- Preserve raw data before calculations or cleaning.
- Document exclusions and corrections instead of deleting their history.
- Store enough method detail that a later reader can reproduce the measurement.
Visual references
Use diagrams to support the literature.
Visuals help with anatomy, comparisons, and measurement concepts, but they do not replace context or diagnosis.
Continue learning
Move sideways only when the evidence calls for it.
Related THC subjects remain one click away without overwhelming the page with another full catalog.