sbh4th
commented
7 years ago The only thing I can think of to explain this is sampling error, which occurred in only this year but I have no idea why it might have affected CA men in particular, and white men at that. NCHS mortality reports routinely add this note when they talk about random variation of mortality rates:
“The mortality data presented in this report, with the exception of data for 1972, are not subject to sampling error. In 1972, mortality data were based on a 50% sample of deaths because of resource constraints” https://www.cdc.gov/nchs/data/nvsr/nvsr65/nvsr65_04.pdf https://www.cdc.gov/nchs/data/nvsr/nvsr65/nvsr65_04.pdf
For this reason I think it’s okay to remove this year if we want, or we can leave it in and have at least a vaguely plausible explanation for why things may look odd in this year. I’m okay either way, but it will need to be explained regardless.
Thanks for letting us know!
Sam
On Feb 1, 2017, at 7:43 AM, Corinne Riddell notifications@github.com wrote:
Hi @jkaufman94 https://github.com/jkaufman94 @sbh4th https://github.com/sbh4th, and cc @kathryntmorrison https://github.com/kathryntmorrison:
For some reason, deaths due to CVD in California Males in 1972 are much lower than contiguous years and as far as I can tell it must be a coding error in the original data. You can see the blip in male life expectancy gap here:
https://cloud.githubusercontent.com/assets/12090129/22512566/6d6d10e6-e867-11e6-925d-709d661251e6.png And when I go to the raw data, you can see that for white males the crude mortality rate for COD for whites (teal line) is lower in 1972 than surrounding years, with no similar affect for blacks (pink line):
https://cloud.githubusercontent.com/assets/12090129/22513307/d626c6b6-e869-11e6-94dc-2a18e4725533.png At the moment, I've included this data and it seems to only affect the calculation of life expectancy for white males in 1972, the BW LE gap for males in 1972 and the contribution of CVD to the LE gap in 1972. It doesn't have a material effect on the COD decomposition for the other causes, but does have an effect on the % contribution of other causes (which is to be expected since the LE gap increases in 1972 because of this supposed decrease in CVD related mortality in white men).
Do you agree with me removing these data (i.e., CVD related deaths in white males in 1972) from the analysis and not reporting white male LE in 1972, the BW gap, or any of the decompositions for males in 1972?
— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/corinne-riddell/BlackWhiteMortalityGap/issues/191, or mute the thread https://github.com/notifications/unsubscribe-auth/AQMMLqE6THSEx7HrJWWZd2soiVh9a2uKks5rYKgigaJpZM4L0BHu.
jkaufman94
corinne-riddell
Hi @jkaufman94 @sbh4th, and cc @kathryntmorrison:
For some reason, deaths due to CVD in California Males in 1972 are much lower than contiguous years and as far as I can tell it must be a coding error in the original data. You can see the blip in male life expectancy gap here:
And when I go to the raw data, you can see that for white males the crude mortality rate for COD for whites (teal line) is lower in 1972 than surrounding years, with no similar affect for blacks (pink line):
At the moment, I've included this data and it seems to only affect the calculation of life expectancy for white males in 1972, the BW LE gap for males in 1972 and the contribution of CVD to the LE gap in 1972. It doesn't have a material effect on the COD decomposition for the other causes, but does have an effect on the % contribution of other causes (which is to be expected since the LE gap increases in 1972 because of this supposed decrease in CVD related mortality in white men).
Do you agree with me removing these data (i.e., CVD related deaths in white males in 1972) from the analysis and not reporting white male LE in 1972, the BW gap, or any of the decompositions for males in 1972?