3 ms·

I think you've seriously gotta squint to make the data you're talking about look like anything other than noise.

ajross 14y ago
That's not right. Correlations with data this large can be proven quite conclusively even if the signal is low. There are a lot of deaths every year, so even fairly small relative signals are resolvable. What's interesting here though is that this is not a small signal, precisely because it shows up against the background without any correlation work at all. FTA: > "We find the largest rise in American mortality rates [in the last 15 years] occurred in 1999, the year Vioxx was introduced, while the largest drop occurred in 2004, the year it was withdrawn," says Unz.
- nkoren 14y ago
  That statement, however, doesn't seem to correspond with the CDC's graph, which shows a much larger rise in death rates occurring in 1993. (Source article for that graph: http://www.cdc.gov/nchs/data/databriefs/db88.htm http://www.cdc.gov/nchs/data/databriefs/db88.htm -- would that they distributed the actual datasets behind the graphs!)
  - ajross 14y ago
    1993 was 19 years ago. Though I'd be curious what that effect was too.
    - DanBC 14y ago
      Flu outbreak? EDIT: URL removed. The link I included appear to be written by an idiot. Sorry. > A CDC review of mortality patterns in 1993 states: “the decline in life expectancy likely reflects increases in death rates for chronic diseases during the two influenza outbreaks in 1993.” [Morbidity Mortality Weekly 45 (08), 161-64, March 1, 1996] Some increases in chronic disease (diabetes, heart disease, COPD) were “the result of the two influenza epidemics of 1993,” said the report. [Monthly Vital Statistics Report, Volume 44, No. 7(S), page 9, Feb. 29, 1996]
    - nkoren 14y ago
      Crikey, that was 19 years ago already? No wonder I've got a stiff neck! Apologies -- I'd misread the OP to say that it had been the largest jump in 15 years -- meaning the 15 years prior to 1999.

That's not right. Correlations with data this large can be proven quite conclusively even if the signal is low. There are a lot of deaths every year, so even fairly small relative signals are resolvable. What's interesting here though is that this is not a small signal, precisely because it shows up against the background without any correlation work at all. FTA: > "We find the largest rise in American mortality rates [in the last 15 years] occurred in 1999, the year Vioxx was introduced, while the largest drop occurred in 2004, the year it was withdrawn," says Unz.

nkoren 14y ago
That statement, however, doesn't seem to correspond with the CDC's graph, which shows a much larger rise in death rates occurring in 1993. (Source article for that graph: http://www.cdc.gov/nchs/data/databriefs/db88.htm http://www.cdc.gov/nchs/data/databriefs/db88.htm -- would that they distributed the actual datasets behind the graphs!)
- ajross 14y ago
  1993 was 19 years ago. Though I'd be curious what that effect was too.
  - DanBC 14y ago
    Flu outbreak? EDIT: URL removed. The link I included appear to be written by an idiot. Sorry. > A CDC review of mortality patterns in 1993 states: “the decline in life expectancy likely reflects increases in death rates for chronic diseases during the two influenza outbreaks in 1993.” [Morbidity Mortality Weekly 45 (08), 161-64, March 1, 1996] Some increases in chronic disease (diabetes, heart disease, COPD) were “the result of the two influenza epidemics of 1993,” said the report. [Monthly Vital Statistics Report, Volume 44, No. 7(S), page 9, Feb. 29, 1996]
  - nkoren 14y ago
    Crikey, that was 19 years ago already? No wonder I've got a stiff neck! Apologies -- I'd misread the OP to say that it had been the largest jump in 15 years -- meaning the 15 years prior to 1999.