search

swish-climate-impact-assessment / BiosmokeValidatedEvents

bio_validated_bushfire_events
http://swish-climate-impact-assessment.github.io/BiosmokeValidatedEvents
2 stars 1 forks source link

Thresholds: 75% of hours per day. 70% days across time series. %5 missing days then avg prior/next day #5

Closed ivanhanigan closed 8 years ago

ivanhanigan commented 8 years ago

@farhadsalimi, https://github.com/swish-climate-impact-assessment/BiosmokeValidatedEvents/blame/master/inst/doc/impute_aphea2_sydney_pm25.Rmd#L167 Talks about selecting the sites which have at least 75% of data available for each variable, but our protocol used 70%.
I'd tried to make this accessible (modifiable) in this function: https://github.com/swish-climate-impact-assessment/BiosmokeValidatedEvents/blob/master/R/sites_todo.R

ivanhanigan commented 8 years ago

@farhadsalimi, also this 5% rule. Comes from http://www.invs.sante.fr/publications/2001/apheis/Apheis_p101-136.pdf p110. So we need to first check before running line 263 https://github.com/swish-climate-impact-assessment/BiosmokeValidatedEvents/blame/master/inst/doc/impute_aphea2_sydney_pm25.Rmd#L263

farhadsalimi commented 8 years ago

@ivanhanigan will do that Ivan, does it mean that if it is more than 5%, then do nothing?

ivanhanigan commented 8 years ago

yes, don't avg pre-or-post days at all

On Thu, Mar 17, 2016 at 12:04 PM, farhadsalimi notifications@github.com wrote:

I will do that Ivan, does it mean that if it is more than 5%, then do nothing?

— You are receiving this because you authored the thread. Reply to this email directly or view it on GitHub https://github.com/swish-climate-impact-assessment/BiosmokeValidatedEvents/issues/5#issuecomment-197629526

farhadsalimi commented 8 years ago

@ivanhanigan i followed them (http://jech.bmj.com/content/50/Suppl_1/S12.full.pdf+html) thats why i used 75% hourly and whole time series. What do you think?

ivanhanigan commented 8 years ago

hmmm I think I am getting confused about which sources we follow. the link above is to Katsouyanni, K., Schwartz, J., Spix, C., Touloumi, G., Zmirou, D., Zanobetti, A., … Anderson, H. R. (1996). Short term effects of air pollution on health: a European approach using epidemiologic time series data: the APHEA protocol. Journal of Epidemiology & Community Health, 50(Suppl 1), S12–S18. doi:10.1136/jech.50.Suppl_1.S12 who use three monthly avgs (but don't say if rolling/centred or calendar season)

I have in my doco from years ago the Touloumi, G., Samoli, E., Le Tetre, A., Atkinson, R., & Schwartz, J. (2001). Annex 1 - APHEA2 Methodology. Air Pollution and Health: a European Information System (APHEIS). Retrieved from http://www.invs.sante.fr/publications/2001/apheis/ which used yearly averages

In our paper (2011) we cite Atkinson, R. W., Anderson, R. H., Sunyer, J., Ayres, J., Baccini, M., Vonk, J. M., … Katsouyanni, K. (2001). Acute Effects of Particulate Air Pollution on Respiratory Admissions. American Journal of Respiratory and Critical Care Medicine, 164(10), 1860–1866. doi:10.1164/ajrccm.164.10.2010138 and ATKINSON, R. W., ROSS ANDERSON, H., SUNYER, J., AYRES, J., BACCINI, M., VONK, J. M., … KATSOUYANNI, K. (2001). Online Supplement for: ACUTE EFFECTS OF PARTICULATE AIR POLLUTION ON RESPIRATORY ADMISSIONS. American Journal of Respiratory and Critical Care Medicine, 164(10). But they do not specify.

I think Fay used that ref for simplicity. In reality the 3mo-rolling-centred approach probably originated in the code that Geoff and Richard Summerhayes.

I reckon we go with your R code that does calendar season, and cite the Katsouyanni paper, but we should also note that this differs to the protocol used in Johnston et al 2011. We can always check later if it makes a difference.

On Thu, Mar 17, 2016 at 12:29 PM, farhadsalimi notifications@github.com wrote:

@ivanhanigan https://github.com/ivanhanigan i followed them (http://jech.bmj.com/content/50/Suppl_1/S12.full.pdf+html) thats why i used 70% hourly and whole time series. What do you think?

— You are receiving this because you were mentioned. Reply to this email directly or view it on GitHub https://github.com/swish-climate-impact-assessment/BiosmokeValidatedEvents/issues/5#issuecomment-197637802

ivanhanigan commented 8 years ago

apology, my comment above was not directly about about thresholds. I see Karsouyanni say 75% of days across time series, but in Johnston 2011 we went with 70% days across the time-series after looking at our local data. I think we can still follow the rule that we cite Katsouyanni for imputation and refer to Johnston 2011 regarding the protocol that I used.

On Thu, Mar 17, 2016 at 12:45 PM, Ivan Hanigan ivan.hanigan@gmail.com wrote:

hmmm I think I am getting confused about which sources we follow. the link above is to Katsouyanni, K., Schwartz, J., Spix, C., Touloumi, G., Zmirou, D., Zanobetti, A., … Anderson, H. R. (1996). Short term effects of air pollution on health: a European approach using epidemiologic time series data: the APHEA protocol. Journal of Epidemiology & Community Health, 50(Suppl 1), S12–S18. doi:10.1136/jech.50.Suppl_1.S12 who use three monthly avgs (but don't say if rolling/centred or calendar season)

I have in my doco from years ago the Touloumi, G., Samoli, E., Le Tetre, A., Atkinson, R., & Schwartz, J. (2001). Annex 1 - APHEA2 Methodology. Air Pollution and Health: a European Information System (APHEIS). Retrieved from http://www.invs.sante.fr/publications/2001/apheis/ which used yearly averages

In our paper (2011) we cite Atkinson, R. W., Anderson, R. H., Sunyer, J., Ayres, J., Baccini, M., Vonk, J. M., … Katsouyanni, K. (2001). Acute Effects of Particulate Air Pollution on Respiratory Admissions. American Journal of Respiratory and Critical Care Medicine, 164(10), 1860–1866. doi:10.1164/ajrccm.164.10.2010138 and ATKINSON, R. W., ROSS ANDERSON, H., SUNYER, J., AYRES, J., BACCINI, M., VONK, J. M., … KATSOUYANNI, K. (2001). Online Supplement for: ACUTE EFFECTS OF PARTICULATE AIR POLLUTION ON RESPIRATORY ADMISSIONS. American Journal of Respiratory and Critical Care Medicine, 164(10). But they do not specify.

I think Fay used that ref for simplicity. In reality the 3mo-rolling-centred approach probably originated in the code that Geoff and Richard Summerhayes.

I reckon we go with your R code that does calendar season, and cite the Katsouyanni paper, but we should also note that this differs to the protocol used in Johnston et al 2011. We can always check later if it makes a difference.

On Thu, Mar 17, 2016 at 12:29 PM, farhadsalimi notifications@github.com wrote:

@ivanhanigan https://github.com/ivanhanigan i followed them (http://jech.bmj.com/content/50/Suppl_1/S12.full.pdf+html ) thats why i used 70% hourly and whole time series. What do you think?

— You are receiving this because you were mentioned. Reply to this email directly or view it on GitHub https://github.com/swish-climate-impact-assessment/BiosmokeValidatedEvents/issues/5#issuecomment-197637802

ivanhanigan commented 8 years ago

@farhadsalimi thanks for sorting the 5% rule. I'm going to start a new issue regarding the number of sites to use as Geoff's instructions are a bit complicated.