Statistical tests: matched pairs categorical dataBMJ 2011; 343 doi: https://doi.org/10.1136/bmj.d6623 (Published 19 October 2011) Cite this as: BMJ 2011;343:d6623
- Philip Sedgwick, senior lecturer in medical statistics
- 1Centre for Medical and Healthcare Education, St George’s, University of London, Tooting, London, UK
Researchers investigated the effectiveness of seat belts for protecting children involved in car crashes.1 They selected a sample of all car crashes within prescribed geographical regions of Canada between 1984 and 1992 that resulted in injury or death. Cars involved in a crash were selected only if one of the occupants was aged 4-14 years and the seat belt status (belted or unbelted) of all occupants was known.
The sample comprised 470 cars involved in a crash. For cars with just one child occupant, the researchers recorded whether the child and driver in the car were wearing seat belts, thereby providing two measurements matched by car occupancy. If a car had more than one child occupant, one of the children was randomly selected. The table⇓ shows the association between the seat belt status of the driver and child for each of the 470 matched pairs. The researchers reported a significant difference between drivers and children in the proportion who were wearing a seat belt (P=0.0001). Overall, 59.8% of children were wearing a seat belt compared to …