Unit 01 · Required Dataset Laboratory
Urban Shade & Surface Temperature
Audit a fictional observational dataset from raw values to defensible inference. Build and critique visualizations, quantify selected differences, identify unusual observations and confounders, and distinguish what the data show from what a causal study would need to establish.
This laboratory uses a fictional dataset and requires no outdoor measurement. Do not interpret it as authorization for unsupervised fieldwork near roads, hot surfaces, construction areas, private property, or other hazards.
A. Study description and design classification
At eight fictional paved sites, a student estimated the percentage of the measurement area shaded at 2:00 p.m. and recorded surface temperature with the same infrared thermometer. The student did not assign shade levels, randomly allocate sites to shade conditions, or fully control differences among locations.
This is best classified as an observational study because ____________________________________. The explanatory variable is __________________; the response variable is __________________.
B. Raw dataset
| Site | Estimated shade (%) | Surface temperature (°C) |
|---|---|---|
| A | 5 | 49.2 |
| B | 15 | 47.8 |
| C | 25 | 45.4 |
| D | 35 | 43.9 |
| E | 50 | 40.7 |
| F | 65 | 37.8 |
| G | 80 | 35.6 |
| H | 90 | 36.4 |
These values are instructional and fictional. They are not measurements of named real locations and should not be cited as real urban-temperature evidence.
C. Data-quality audit
Before graphing, inspect what the table does—and does not—tell you.
- Are units supplied for both measured/estimated quantities?
- Is shade a direct instrument measurement or an estimate?
- Does “same infrared thermometer” guarantee identical measurement conditions? Explain.
- Which important metadata are missing—for example surface material, exact measurement protocol, weather, prior sun exposure, moisture, or repeated readings?
- Why would entering a missing temperature as 0 °C be scientifically incorrect?
D. Visualization laboratory
Create a scatter plot with estimated shade (%) on the x-axis and surface temperature (°C) on the y-axis. Plot all eight sites and use a descriptive title.
Title: ____________________. X-axis range/interval: ____________________. Y-axis range/interval: ____________________. Will you connect the points? ______. Defend that decision.
Because these are separate sites rather than sequential observations along one continuous path, connecting A→B→C→… can imply a continuity or sequence that the study did not measure. A scatter plot can show the association without requiring such a connection.
E. Scale integrity challenge
Make or imagine two defensible versions of the same scatter plot: one with a broad y-axis that provides more absolute context and one with a narrower y-axis that makes differences easier to inspect.
Which version makes the temperature differences appear more dramatic? Which better shows their magnitude in a wider context? Does either change the underlying numbers? State what axis information a reader needs before interpreting visual steepness.
F. Quantitative checks
- Calculate the temperature difference between Site A and Site G: ______ °C.
- Calculate the temperature difference between Site G and Site H: ______ °C.
- State the observed shade range: ______% to ______%.
- State the observed temperature range: ______ °C to ______ °C.
- Which site departs most visibly from a strictly decreasing sequence at the high-shade end? ______.
Use the Khaemenes Scientific Calculator to verify arithmetic if desired, but show enough work that the result can be audited.
G. Describe the pattern before explaining it
Describe direction, scope, and the high-shade departure using actual values. Do not use “because” and do not claim cause yet.
H. Anomaly reasoning
Site H does not continue the otherwise decreasing sequence from A through G. That does not make H an “error.”
Give at least three scientifically plausible explanations for the departure. For each, identify what additional information or measurement would help test it.
Possible categories to consider include ordinary site-to-site variation, shade-estimation uncertainty, surface material, moisture, airflow, prior solar exposure, measurement geometry, or a real nonlinear relationship. Do not choose among them without evidence.
I. Confounder map
A useful confounder must plausibly relate to both shade and surface temperature, not merely be another fact about the site.
| Possible variable | How it could relate to shade | How it could affect temperature | What evidence would help? |
|---|---|---|---|
| Surface material | |||
| Moisture | |||
| Prior sun exposure / time shaded | |||
| Your own candidate |
J. Association versus causation
Write the strongest claim justified directly by this observational dataset.
Write a stronger causal claim that these observations alone do not establish, then explain exactly why the design cannot isolate that cause.
The dataset supports a negative association in these fictional observations. It does not establish that shade was the only cause—or quantify the isolated causal effect of shade—because shade was not assigned and relevant site differences were not controlled.
K. Interpolation and extrapolation boundary
The observed shade values span 5%–90%. Any prediction outside the observed range requires extrapolation and additional assumptions.
Would estimating temperature near 60% shade be interpolation or extrapolation? What about predicting the temperature of a 100% shaded site? Explain why neither prediction is guaranteed even if a trend appears strong.
L. Controlled follow-up design
Propose a study that better isolates the effect of shade. A strong design should reduce site differences rather than merely collect more of the same confounded observations.
- State the causal question.
- Identify what would be deliberately manipulated or carefully matched.
- Define the temperature measurement method and timing.
- Control or measure surface material, moisture, exposure history, measurement geometry, and relevant weather conditions.
- Include replication and a condition-order/site-selection strategy where appropriate.
- State one remaining limitation.
M. Optional extension: external evidence
If you compare this fictional exercise with real research about shade, urban heat, or surface temperature, use Evidence Citation Studio to keep external evidence clearly separate from this dataset. Record source, method, relevant finding, limitation, and how the evidence relates to—but does not retroactively transform—the fictional observations.
N. Mentor causal-defense conference
Bring your graph, quantitative checks, Site H analysis, confounder map, supported claim, and follow-up design to the Science Mentor. Be prepared to answer:
- What does the graph show before you explain why?
- Why is Site H not automatically an error?
- Which confounder is most threatening and why?
- What causal sentence did you refuse to make?
- What design change would most increase causal confidence?
Mentor and Science Studio support checking and revision. The student remains responsible for analysis and scientific claims.
O. Evidence-complete gate
Before recording completion, verify that you have a labeled scatter plot, scale audit, quantitative checks, pattern description, anomaly analysis, confounder map, supported association claim, unsupported-causal-claim explanation, inference boundary, and controlled follow-up design.