Test explanations with controls

In November 2017, the DAMPE collaboration reported a break near 0.9 teraelectronvolts in the combined electron-and-positron spectrum. The result constrained models of high-energy particles. The team discussed possible astrophysical sources and dark-matter interpretations; the measurement did not uniquely establish dark matter.

A spectrum describes how measured particle flux changes with energy. A change in slope can be a real feature while its cause remains uncertain. Ask which competing explanations predict the feature, whether they also fit other observations and whether instrument effects were excluded. Naming a mission after a scientific question does not preselect the answer.

Shijian-10’s 2016 mission included experiments in fluids, materials and life science. Its recoverable capsule brought samples back for laboratory analysis after twelve days in orbit. The National Space Science Data Center describes experiments with comparisons to ground samples. Physical return made additional measurements possible; it was not itself a conclusion about a biological mechanism.

In an original experiment, flight samples and ground controls differ in both radiation exposure and temperature history. If their outcomes differ, assigning the entire change to microgravity would be premature. A stronger design records these conditions, matches what can be matched and uses additional controls or measurements to distinguish explanations.

Suppose an invented material experiment measures average growth of 12 units in a flight group and 10 units in a matched control group. The difference is 2 units, or 20 percent of the control mean. These values illustrate arithmetic, not a Shijian-10 result. Without sample size and variation, they do not establish how precise or repeatable the effect is.

A useful result includes independent samples, the treatment conditions, uncertainty and the analysis choices. Many readings from one specimen are not automatically many independent specimens. Predefined comparisons and independent repetition make a claim easier to assess, especially when investigators can examine many possible outcomes.

Separate observation, comparison and explanation