A suggestion for the US agencies involved with investigating the Havana syndrome: take the subset of employees affected who objectively visible brain damage (n

A suggestion for the US agencies involved with investigating the Havana syndrome: take the subset of employees affected who objectively visible brain damage (not to separate the "false" and "true" victims, but to enrich that fraction for investigation purpouses, alternatively accept all self-reported victims with potentially lower signal to noise ratio as a result). next take a subset of unaffected employees with similar task distributions as a control group then for each person (affected or control) check the origins of all mail, check scheduled appointments with Iranian nationals, ask if they recall ever experiencing something being spritzed close to their face (they may have interpreted as something innocuous like perfume), allow them to make as many entries as possible, then ask the same question (any dates or times of occurence, especially before onset of symptoms) but for specific odors (cherry odor = control, vanilla odor = control, bitter almond odor = scent of hydrogen cyanide, mint odor = control, eye irritant = added to disinfection hydrogen cyanide for repelling human end-users, ...), then for each type you take the difference between time of symptom onset minus odor exposure event and convert to same units (floating point days for example plus or minus estimated error), convert these time intervals to reciprocals for rates ( floating point "per day"), then for each type of odor, or for each type of meeting with Iranian, or for each parcel from Iranian nationals, make a grand sum of that rate once for the affected group, and once for the unaffected control group. then compare the mean rates between control group and affected group, and use hypothesis testing statistics to calculate the likelihood for the observed rates in the affected group given the rates from the unaffected group, if the null hypothesis of mere random clustering can be rejected you have detected that it is unlikely to explain by random chance, and those Iran associated links are proven significant (not in size of correlation, but that the correlation however small or large can not be explained away by chance). You could redo (or simultaneously do the same exercise with different nationalities).