Lesson 6 of 6 · Testing an interpretation against uncertainty
Make a claim that can survive another observation
10 minute lesson · 3 checks along the way
Your goal for this lessonCompare competing explanations with numerical uncertainty, calibration checks, and a purposeful follow-up observation.
Follow a linked name to explore its archive card. “Back to your lesson” brings you to the same part of the reading.
Words with a dotted underline open a short definition right here in the lesson.
A reported number has a measurement history
An astronomical result is more than a best estimate. It includes what was measured, how it was calibrated, the assumptions connecting it to a physical quantity, and its uncertainty. Random variation can scatter repeated measurements. A systematic effect, such as an incorrect sensitivity , can shift many measurements together. More exposures often reduce some random contributions, but repeating the same biased setup does not automatically remove a shared calibration error.
Precision describes how tightly a method constrains or repeats a value; accuracy concerns agreement with the value being measured. Neither is established by printing more decimal places. Uncertainty intervals also need definitions. A is not a guaranteed maximum error, and a confidence statement depends on its statistical procedure. This course uses simplified comparisons to expose the logic, not a universal rule that a particular number of uncertainty units proves a discovery.
Check 1 of 3
10 pointsUse the reading above. You can try again without losing points.
Work through a decision
Compare the size of a difference with its uncertainty
Two fictional, independent calibrated measurements of the same stable source in the same band are 100 ± 6 and 112 ± 8 units. Suppose the uncertainties are one-standard-deviation random errors, approximately Gaussian, with no unaccounted systematic contribution. The difference is 12 units. For independent errors, the standard uncertainty of the difference is √(6² + 8²) = 10 units. The discrepancy is therefore only 1.2 standard deviations.
That comparison alone gives little reason to declare the source variable. It also does not prove perfect constancy. If the two measurements share a calibration error, the independence assumption needs revision; correlations affect the uncertainty of their difference. The quoted error bars should describe what was included. A small formal error from a fitting routine can be misleading when the model or calibration omits a relevant effect.
Check 2 of 3
10 pointsUse the reading above. You can try again without losing points.
Design a follow-up that could change your mind
Imagine a faint brightening appears only in a detector corner. One explanation is a celestial transient; another is a local detector defect. A repeat observation with the sky moved onto different detector pixels can test the distinction. A real sky source should remain associated with the same sky coordinates after registration, while a fixed detector feature follows detector coordinates. Timing matters because a real transient may also fade between exposures.
Independent instruments, comparison stars, additional wavelengths, and laboratory calibration each probe different weaknesses. Choose the test according to the competing predictions. Measuring a thousand copies of the same processing artifact does not create a thousand independent confirmations. Likewise, selecting only the most striking fluctuations from a huge survey makes apparently unusual events easier to find by chance. The search procedure belongs in the evidence assessment.
Read an engineering observation within its purpose
The Spitzer Space Telescope’s early aliveness-test image shows infrared detections during instrument checkout. NASA describes a 100-second observation made before optimal temperature and focus were reached. It established useful evidence that the observatory was detecting celestial signals; it was not a claim that every operating condition had already reached its final science state. The bright spots and display colors should be interpreted alongside that engineering context.
End a study with a claim scaled to its evidence: what is supported, which alternatives remain, and what observation would distinguish them. Report a as a sensitivity-dependent limit when that is what the analysis provides. Preserve the methods and calibrated measurements so another investigator can examine the result. Astronomers learn about inaccessible places by making the path from observation to interpretation explicit, testable, and open to revision.
The Spitzer Space Telescope’s PIA04724 aliveness-test image uses infrared observations displayed in visible colors. NASA describes 100 seconds of observing time during September 2003 checkout, before optimal operating temperature and focus were reached. It is an engineering observation of celestial sources, not a photograph of the spacecraft or a map of naked-eye colors.
Look closely
What useful conclusion can a checkout image support before every science operating condition is optimized?
See what to notice
It can demonstrate that the instrument detects celestial signals and test aspects of its operation. Precision science measurements still require the appropriate calibration, response, focus and temperature assessment, and uncertainty analysis. An early successful detection is not evidence that all later performance goals have already been met.
Check 3 of 3
10 pointsUse the reading above. You can try again without losing points.
Put it together
Your lesson progress
Each check connects to a part of the story. Complete all 3 to earn your 25-point lesson bonus.
Go to the source
Explore the evidence behind this lesson. These optional readings and videos open in a new tab.