"Risk" is the likelihood of one thing and not another thing happening. "Relative risk" and "absolute risk" are types of risk that a smart science writer differentiates and reports in a non-biased way.
Statistics can be used deceptively in reporting science results by confusing relative risk with absolute risk.
One can report that "taking an aspirin quadruples stroke risk for that day", which sounds dramatic. This report of relative risk makes aspirin sound very dangerous. But what if we look at absolute riskl. The actual numbers are stroke without aspirin are .00000000001 per cent likelihood for that day and stroke with aspirin is 000000000000004 per cent for that day. Yes, the risk has quadrupled but the real difference is negligible.
RELATIVE RISK
Statistics can be used deceptively in reporting science results by confusing relative risk with absolute risk.
One can report that "taking an aspirin quadruples stroke risk for that day", which sounds dramatic. This report of relative risk makes aspirin sound very dangerous. But what if we look at absolute riskl. The actual numbers are stroke without aspirin are .00000000001 per cent likelihood for that day and stroke with aspirin is 000000000000004 per cent for that day. Yes, the risk has quadrupled but the real difference is negligible.
RELATIVE RISK
So, again, what is relative risk. Consider this somewhat whimsical example of relative risk. A researcher finds that 10 of 100 boys age 12 want a unicorn poster. 40 of 100 girls age 12 want a unicorn poster. The researcher writes, “12 year girls are four times more likely than age adjusted boys to want to unicorn poster.”
He can also report the relative risk another way--by percentile. “Being female increases the likelihood by 400 per cent the number of 12 year olds wishing to have a unicorn poster.
A slightly different third option in reporting relative risk is not to use the multiplicative but additive function. “Female rather than male gender increases by a sum of 300% the number of pre-adolescents desirous of a poster depicting a unicorn.” How is the 300% calculated? The quantity of poster lovers in the female group was 30 more than in the male group, and 30 is 300% of 10.
MISLEADING USE OF RELATIVE RISK
Take this made up study to see how relative risk can be misused. Say that a study of archival hospital records since 1900 reveal that an Irish citizen had a 1 in 1000 billion likelihood of being born with two heads. The records reveal that a French citizen has a 4 in 1000 billion chance of being born with two heads. The researcher reports relative risk, “Our exhaustive examination of records showed that French rather than Irish citizenship quadruples the risk of having an extra head. .” Or, she can say the same thing using a percentile. “A thorough study of hospital records indicates that French rather than Irish citizenship multiplies by 400% the likelihood of having an extra head.
ABSOLUTE RISK
Absolute Risk is the overall likelihood of an event occurring in a population. When absolute risk is extremely low, a report of relative risk may mislead, as illustrated by the overdramatic report of the centaur research. Despite the researcher’s emphatic claim, citizenship only infinitesimally alters the likelihood of being born with an extra head. Absolute risk for centaur birth remains far out on the asymptote, a likelihood less than 1 in 100 million.
Johns Hopkins Science Writing Program
No comments:
Post a Comment