A research poster from the School of Underlying Conditions works out how much a sample still says once every reading has been booked to the nearest mark
An exact calculation shows that what a graduated instrument keeps of a sample depends on where its marks happen to fall
How much is lost when a reading is written down to the nearest mark? The question has an old and serviceable answer, and a poster released today by the School of Underlying Conditions reports where that answer stops being the whole of it. Informative Only at the Mark: Fisher Information in a Normal Sample Read on a Coarse Scale is the work of Dr Brynjar Obiora, Senior Research Fellow, and Dr Vigdis Emenike, Senior Lecturer.
The poster finds that while an instrument’s divisions are no wider than the scatter of what it measures, it makes no difference where the marks sit. Once they are wider, it makes nearly all the difference. A quantity whose average lies on a mark keeps close to two-thirds of what the sample had to say about it, however coarse the scale becomes; one lying midway between two marks keeps, in the end, nothing. What the sample says about its own spread behaves the opposite way.
I have read a tape to the nearest mark for most of my working life and it had not occurred to me that the mark might have a view. I cannot say the tape reads any differently this week, but I notice that I hold it with more respect.
— Dr Vigdis Emenike, Senior Lecturer
The work proceeds entirely by calculation, bringing the School’s field practice into conversation with a long-established strand of statistical theory and carrying the result through to a recommendation about pairing instruments. The University counts it among the quieter satisfactions of an earth sciences school that its equipment can occasion a result without being taken out of the shed.
The full poster is available from the University’s research repository under an open licence, doi:10.5555/slop.wk8llm.