Heat wave and Ozone, Mediation analysis on mortality (Meteo France), test for non linearity

Temperature Data Management

Creating Database only with summer months

villes_s<-list()

for(i in 1:length(villes)) {
  villes_s[[i]]<-villes[[i]][villes[[i]]$saison=="summer",]
  villes_s[[i]]$temps <- ave(villes_s[[i]]$time,villes_s[[i]]$annee,  FUN = seq_along)
  }

names(villes_s)<-cities

table of N obs per months to check to have only summer month

for (i in 1:length(villes)){
print(table(villes_s[[i]]$mois))
}

## 
##   1   2   3   4   5   6   7   8   9  10  11  12 
##   0   0   0   0   0 480 496 496 480   0   0   0 
## 
##   1   2   3   4   5   6   7   8   9  10  11  12 
##   0   0   0   0   0 480 496 496 480   0   0   0 
## 
##   1   2   3   4   5   6   7   8   9  10  11  12 
##   0   0   0   0   0 480 496 496 480   0   0   0 
## 
##   1   2   3   4   5   6   7   8   9  10  11  12 
##   0   0   0   0   0 480 496 496 480   0   0   0 
## 
##   1   2   3   4   5   6   7   8   9  10  11  12 
##   0   0   0   0   0 480 496 496 480   0   0   0 
## 
##   1   2   3   4   5   6   7   8   9  10  11  12 
##   0   0   0   0   0 480 496 496 480   0   0   0 
## 
##   1   2   3   4   5   6   7   8   9  10  11  12 
##   0   0   0   0   0 480 496 496 480   0   0   0 
## 
##   1   2   3   4   5   6   7   8   9  10  11  12 
##   0   0   0   0   0 480 496 496 480   0   0   0 
## 
##   1   2   3   4   5   6   7   8   9  10  11  12 
##   0   0   0   0   0 480 496 496 480   0   0   0 
## 
##   1   2   3   4   5   6   7   8   9  10  11  12 
##   0   0   0   0   0 480 496 496 480   0   0   0 
## 
##   1   2   3   4   5   6   7   8   9  10  11  12 
##   0   0   0   0   0 480 496 496 480   0   0   0 
## 
##   1   2   3   4   5   6   7   8   9  10  11  12 
##   0   0   0   0   0 480 496 496 480   0   0   0 
## 
##   1   2   3   4   5   6   7   8   9  10  11  12 
##   0   0   0   0   0 480 496 496 480   0   0   0 
## 
##   1   2   3   4   5   6   7   8   9  10  11  12 
##   0   0   0   0   0 480 496 496 480   0   0   0 
## 
##   1   2   3   4   5   6   7   8   9  10  11  12 
##   0   0   0   0   0 480 496 496 480   0   0   0

Creating Heat Wave variable

## First condition for heat wave
for (i in 1:length(villes_s)){
  villes_s[[i]]$heat_wave1<-NA
  for (r in 3:nrow(villes_s[[i]])){
  villes_s[[i]]$heat_wave1[r]<-  if (villes_s[[i]]$tempmoy[r] >= trshld975[i] & villes_s[[i]]$tempmoy[r-1] >= trshld975[i] & villes_s[[i]]$tempmoy[r-2] >= trshld975[i] ) 1 else 0
  }}

## Second condition for heat wave
for (i in 1:length(villes_s)){
  for (r in 3:nrow(villes_s[[i]])){
  villes_s[[i]]$heat_wave[r]<-  if (villes_s[[i]]$heat_wave1[r] == 1 & (villes_s[[i]]$tempmoy[r] >= trshld995[i] | villes_s[[i]]$tempmoy[r-1] >= trshld995[i] | villes_s[[i]]$tempmoy[r-2] >= trshld995[i])) 1 else 0
  }}

Number of Heat Wave for each city, for the period from 2000 to 2015:

	Heat wave=0	Heat wave=1
bordeaux	1924	26
clermont	1926	24
grenoble	1914	36
lehavre	1927	23
lyon	1918	32
marseille	1915	35
montpellier	1922	28
nancy	1918	32
nantes	1926	24
nice	1924	26
paris	1923	27
rennes	1918	32
rouen	1925	25
strasbourg	1912	38
toulouse	1921	29

Creating Ozone binary variable

Number of O3>90 perc episodes per city:

	O3>90perc=0	O3>90perc=1
bordeaux	1756	196
clermont	1756	196
grenoble	1756	196
lehavre	1756	196
lyon	1756	196
marseille	1757	195
montpellier	1756	196
nancy	1756	196
nantes	1756	196
nice	1757	195
paris	1756	196
rennes	1756	196
rouen	1756	196
strasbourg	1757	195
toulouse	1756	196

Analysis On Non-Accidental Mortality

Continuous Ozone

Code:

# some data management
 for (i in 1:length(villes_s)){
   villes_s[[i]]$time<-1:nrow(villes_s[[i]]) # Create a new variable for time 
 } 
   

    
  # start bootstrap
     
  nreps=800 #number of bootstrap reps
  results<-matrix(NA,nrow = 15,ncol=10)
  colnames(results)<-c("city","NDE","NDE_2.5","NDE_97.5","NIE","NIE_2.5","NIE_97.5","PMM","PMM_2.5","PMM_97.5")
  
for (i in (1:length(villes_s))){
  boot.nde<- rep(NA,nreps)
  boot.nie <- rep(NA,nreps)
  boot.pmm <- rep(NA,nreps)
  for (rep in 1:nreps)
      
    {
      dat_resample <- sample(1:nrow(villes_s[[i]]),nrow(villes_s[[i]]),replace=T)
      dat_boot <- villes_s[[i]][dat_resample,]
        
      # outcome model - Poisson Regression: Mortality, HW and O3
      m.outcome<-glm(nocc_tot~heat_wave+o3+ns(time,df=round(3*length(time)/122))+Jours+Vacances+hol,data=dat_boot,family=quasipoisson)
      
      # mediator model - multinomial linear regression
      m.mediator<-lm(o3~heat_wave+no2moy+ns(time,df=round(3*length(time)/122))+Jours+Vacances+hol,data=dat_boot) 
      
      # save coefficients
      theta1<-coef(m.outcome)[2]
      
      theta2<-coef(m.outcome)[3]
     
      beta1<-coef(m.mediator)[2]
      
      # estimate CDE and NIE
      
      boot.nde[rep] <- exp(theta1) #NDE
      
      boot.nie[rep] <-exp(theta2*beta1) #NIE 
      
      boot.pmm[rep] <-boot.nde[rep]*(boot.nie[rep]-1)/(boot.nde[rep]*boot.nie[rep]-1) #PMM 
                   
      
    } #end bootstrap

  results[i,1]<-cities[[i]]
  results[i,2]<-median(boot.nde, na.rm=T)
  results[i,3]<-quantile(boot.nde, 0.025, na.rm=T)
  results[i,4]<-quantile(boot.nde, 0.975, na.rm=T)
  results[i,5]<-median(boot.nie, na.rm=T)
  results[i,6]<-quantile(boot.nie, 0.025, na.rm=T)
  results[i,7]<-quantile(boot.nie, 0.975, na.rm=T)
  results[i,8]<-median(boot.pmm, na.rm=T)
  results[i,9]<-quantile(boot.pmm, 0.025, na.rm=T)
  results[i,10]<-quantile(boot.pmm, 0.975, na.rm=T)
       }
  
  
  # output CDE and NIE and confidence intervals

Binary Ozone

Code:

  # start bootstrap
     
  nreps=800 #number of bootstrap reps
  results_bin<-matrix(NA,nrow = 15,ncol=10)
  colnames(results_bin)<-c("city","NDE","NDE_2.5","NDE_97.5","NIE","NIE_2.5","NIE_97.5","PMM","PMM_2.5","PMM_97.5")
  
for (i in (1:length(villes_s))){
  boot.nde<- rep(NA,nreps)
  boot.nie <- rep(NA,nreps)
  boot.pmm <- rep(NA,nreps)
  for (rep in 1:nreps)
      
    {
      dat_resample <- sample(1:nrow(villes_s[[i]]),nrow(villes_s[[i]]),replace=T)
      dat_boot <- villes_s[[i]][dat_resample,] 
        
      # outcome model - Poisson Regression: Mortality, HW and O3
      m.outcome<-glm(nocc_tot~heat_wave+o3_90+ns(time,df=round(3*length(time)/122))+Jours+Vacances+hol,data=dat_boot,family=quasipoisson)
      
      # mediator model - multinomial linear regression
      m.mediator<-lm(o3~heat_wave+no2moy+ns(time,df=round(3*length(time)/122))+Jours+Vacances+hol,data=dat_boot) 
      
      # save coefficients
      theta1<-coef(m.outcome)[2]
      
      theta2<-coef(m.outcome)[3]
     
      beta1<-coef(m.mediator)[2]
      
      # estimate CDE and NIE
      
      boot.nde[rep] <- exp(theta1) #NDE
      
      boot.nie[rep] <-exp(theta2*beta1) #NIE 
      
      boot.pmm[rep] <-boot.nde[rep]*(boot.nie[rep]-1)/(boot.nde[rep]*boot.nie[rep]-1) #PMM 
                   
      
    } #end bootstrap

  results_bin[i,1]<-cities[[i]]
  results_bin[i,2]<-median(boot.nde, na.rm=T)
  results_bin[i,3]<-quantile(boot.nde, 0.025, na.rm=T)
  results_bin[i,4]<-quantile(boot.nde, 0.975, na.rm=T)
  results_bin[i,5]<-median(boot.nie, na.rm=T)
  results_bin[i,6]<-quantile(boot.nie, 0.025, na.rm=T)
  results_bin[i,7]<-quantile(boot.nie, 0.975, na.rm=T)
  results_bin[i,8]<-median(boot.pmm, na.rm=T)
  results_bin[i,9]<-quantile(boot.pmm, 0.025, na.rm=T)
  results_bin[i,10]<-quantile(boot.pmm, 0.975, na.rm=T)
       }
  
  
  # output CDE and NIE and confidence intervals

Results

Results For Continuous Ozone:

city	NDE	NDE_2.5	NDE_97.5	NIE	NIE_2.5	NIE_97.5	PMM	PMM_2.5	PMM_97.5
bordeaux	1.5059	1.2886	1.7479	1.0276	1.0044	1.0549	0.0758	0.0120	0.1560
clermont	1.4536	1.2305	1.7438	1.0181	0.9922	1.0502	0.0552	-0.0249	0.1641
grenoble	1.1647	1.0264	1.3217	1.0321	1.0095	1.0581	0.1845	0.0517	0.5914
lehavre	1.2822	1.0348	1.5426	1.0317	0.9870	1.0837	0.1282	-0.0637	0.5612
lyon	1.3798	1.1874	1.6187	1.0324	1.0166	1.0516	0.1054	0.0522	0.2088
marseille	1.1492	1.0366	1.2496	1.0045	0.9996	1.0126	0.0339	-0.0026	0.1167
montpellier	1.1000	0.9299	1.2816	1.0075	0.9934	1.0235	0.0616	-0.3205	0.7226
nancy	1.2492	1.0689	1.4631	0.9989	0.9787	1.0232	-0.0048	-0.1434	0.1407
nantes	1.2098	1.0231	1.4676	1.0688	1.0278	1.1145	0.2901	0.1110	0.7687
nice	1.2238	1.0513	1.3998	1.0041	0.9984	1.0149	0.0226	-0.0108	0.1225
paris	1.7157	1.4008	2.1435	1.0511	1.0312	1.0821	0.1111	0.0680	0.1658
rennes	1.2876	1.0893	1.5107	1.0548	1.0067	1.1114	0.2027	0.0230	0.4782
rouen	1.3871	1.2058	1.6102	1.0664	1.0281	1.1079	0.1917	0.0864	0.3304
strasbourg	1.4538	1.3017	1.6112	1.0227	1.0019	1.0496	0.0690	0.0052	0.1552
toulouse	1.2413	1.1007	1.3838	1.0333	1.0131	1.0585	0.1493	0.0569	0.3111

Results For Binary Ozone:

city	NDE	NDE_2.5	NDE_97.5	NIE	NIE_2.5	NIE_97.5	PMM	PMM_2.5	PMM_97.5
bordeaux	1.4898	1.2908	1.7319	8.9659	2.5989	41.4950	0.9606	0.8343	0.9926
clermont	1.4086	1.1966	1.6779	8.2179	2.0807	51.7838	0.9617	0.7799	0.9951
grenoble	1.1759	1.0366	1.3386	3.3319	0.6802	17.9409	0.9461	0.1979	1.4579
lehavre	1.3147	1.0697	1.6148	2.5642	0.2251	24.8367	0.9348	-1.6233	3.0791
lyon	1.4056	1.1916	1.6575	2.4020	0.7954	7.3337	0.8368	-0.5225	0.9758
marseille	1.1490	1.0471	1.2523	1.1894	0.9801	1.6733	0.6092	-0.1111	0.8782
montpellier	1.1044	0.9343	1.2851	1.1873	0.6129	2.4631	0.8209	-1.3505	3.4000
nancy	1.2417	1.0633	1.4514	1.1643	0.3542	3.8211	0.8220	-3.6223	5.1768
nantes	1.2410	1.0478	1.4735	25.6910	2.4893	359.7313	0.9929	0.8659	0.9998
nice	1.2100	1.0464	1.3894	1.4231	0.9286	2.9250	0.7282	-0.0964	0.9781
paris	1.7279	1.3741	2.1536	15.2519	5.9191	50.6905	0.9712	0.9273	0.9909
rennes	1.2601	1.0633	1.4940	33.5869	2.8043	703.5327	0.9940	0.8863	0.9998
rouen	1.3990	1.2104	1.6335	32.7542	4.2233	394.9961	0.9911	0.9183	0.9993
strasbourg	1.4281	1.2823	1.5834	7.8511	1.9582	35.4654	0.9576	0.7534	0.9928
toulouse	1.2558	1.1202	1.4032	3.1359	1.0964	10.3081	0.9151	0.3947	0.9831

Forestplot

Natural Direct Effect

Natural Indirect Effect

Analysis On Cardiovascual Mortality

Continuous Ozone

Code:

# some data management
 for (i in 1:length(villes_s)){
   villes_s[[i]]$time<-1:nrow(villes_s[[i]]) # Create a new variable for time 
 } 
   

  # start bootstrap
     
  nreps=800 #number of bootstrap reps
  results<-matrix(NA,nrow = 15,ncol=10)
  colnames(results)<-c("city","NDE","NDE_2.5","NDE_97.5","NIE","NIE_2.5","NIE_97.5","PMM","PMM_2.5","PMM_97.5")
  
for (i in (1:length(villes_s))){
  boot.nde<- rep(NA,nreps)
  boot.nie <- rep(NA,nreps)
  boot.pmm <- rep(NA,nreps)
  for (rep in 1:nreps)
      
    {
      dat_resample <- sample(1:nrow(villes_s[[i]]),nrow(villes_s[[i]]),replace=T)
      dat_boot <- villes_s[[i]][dat_resample,]
        
      # outcome model - Poisson Regression: Mortality, HW and O3
      m.outcome<-glm(cv_tot~heat_wave+o3+ns(time,df=round(3*length(time)/122))+Jours+Vacances+hol,data=dat_boot,family=quasipoisson)
      
      # mediator model - multinomial linear regression
      m.mediator<-lm(o3~heat_wave+no2moy+ns(time,df=round(3*length(time)/122))+Jours+Vacances+hol,data=dat_boot) 
      
      # save coefficients
      theta1<-coef(m.outcome)[2]
      
      theta2<-coef(m.outcome)[3]
     
      beta1<-coef(m.mediator)[2]
      
      # estimate CDE and NIE
      
      boot.nde[rep] <- exp(theta1) #NDE
      
      boot.nie[rep] <-exp(theta2*beta1) #NIE 
      
      boot.pmm[rep] <-boot.nde[rep]*(boot.nie[rep]-1)/(boot.nde[rep]*boot.nie[rep]-1) #PMM 
                   
      
    } #end bootstrap

  results[i,1]<-cities[[i]]
  results[i,2]<-median(boot.nde, na.rm=T)
  results[i,3]<-quantile(boot.nde, 0.025, na.rm=T)
  results[i,4]<-quantile(boot.nde, 0.975, na.rm=T)
  results[i,5]<-median(boot.nie, na.rm=T)
  results[i,6]<-quantile(boot.nie, 0.025, na.rm=T)
  results[i,7]<-quantile(boot.nie, 0.975, na.rm=T)
  results[i,8]<-median(boot.pmm, na.rm=T)
  results[i,9]<-quantile(boot.pmm, 0.025, na.rm=T)
  results[i,10]<-quantile(boot.pmm, 0.975, na.rm=T)
       }
  
  
  # output CDE and NIE and confidence intervals

Binary Ozone

Code:

  # start bootstrap
     
  nreps=800 #number of bootstrap reps
  results_bin<-matrix(NA,nrow = 15,ncol=10)
  colnames(results_bin)<-c("city","NDE","NDE_2.5","NDE_97.5","NIE","NIE_2.5","NIE_97.5","PMM","PMM_2.5","PMM_97.5")
  
for (i in (1:length(villes_s))){
  boot.nde<- rep(NA,nreps)
  boot.nie <- rep(NA,nreps)
  boot.pmm <- rep(NA,nreps)
  for (rep in 1:nreps)
      
    {
      dat_resample <- sample(1:nrow(villes_s[[i]]),nrow(villes_s[[i]]),replace=T)
      dat_boot <- villes_s[[i]][dat_resample,] 
        
      # outcome model - Poisson Regression: Mortality, HW and O3
      m.outcome<-glm(nocc_tot~heat_wave+o3_90+ns(time,df=round(3*length(time)/122))+Jours+Vacances+hol,data=dat_boot,family=quasipoisson)
      
      # mediator model - multinomial linear regression
      m.mediator<-lm(o3~heat_wave+no2moy+ns(time,df=round(3*length(time)/122))+Jours+Vacances+hol,data=dat_boot) 
      
      # save coefficients
      theta1<-coef(m.outcome)[2]
      
      theta2<-coef(m.outcome)[3]
     
      beta1<-coef(m.mediator)[2]
      
      # estimate CDE and NIE
      
      boot.nde[rep] <- exp(theta1) #NDE
      
      boot.nie[rep] <-exp(theta2*beta1) #NIE 
      
      boot.pmm[rep] <-boot.nde[rep]*(boot.nie[rep]-1)/(boot.nde[rep]*boot.nie[rep]-1) #PMM 
                   
      
    } #end bootstrap

  results_bin[i,1]<-cities[[i]]
  results_bin[i,2]<-median(boot.nde, na.rm=T)
  results_bin[i,3]<-quantile(boot.nde, 0.025, na.rm=T)
  results_bin[i,4]<-quantile(boot.nde, 0.975, na.rm=T)
  results_bin[i,5]<-median(boot.nie, na.rm=T)
  results_bin[i,6]<-quantile(boot.nie, 0.025, na.rm=T)
  results_bin[i,7]<-quantile(boot.nie, 0.975, na.rm=T)
  results_bin[i,8]<-median(boot.pmm, na.rm=T)
  results_bin[i,9]<-quantile(boot.pmm, 0.025, na.rm=T)
  results_bin[i,10]<-quantile(boot.pmm, 0.975, na.rm=T)
       }
  
  
  # output CDE and NIE and confidence intervals

Results

Results For Continuous Ozone:

city	NDE	NDE_2.5	NDE_97.5	NIE	NIE_2.5	NIE_97.5	PMM	PMM_2.5	PMM_97.5
bordeaux	1.5333	1.2879	1.8535	1.0275	0.9912	1.0704	0.0739	-0.0246	0.1949
clermont	1.7926	1.3603	2.3460	0.9962	0.9493	1.0481	-0.0088	-0.1381	0.1154
grenoble	1.2618	0.9899	1.5835	1.0561	1.0117	1.1022	0.2107	0.0144	0.9073
lehavre	1.2471	0.8745	1.7303	1.1107	1.0178	1.2255	0.3460	-0.8233	2.6958
lyon	1.4485	1.1374	1.7946	1.0328	1.0055	1.0635	0.0987	0.0164	0.2424
marseille	1.0635	0.9270	1.2312	1.0059	0.9994	1.0178	0.0562	-0.6345	0.9293
montpellier	1.0789	0.7463	1.5156	1.0089	0.9842	1.0408	0.0266	-0.6698	1.0073
nancy	1.5422	1.1576	2.0306	0.9915	0.9508	1.0389	-0.0256	-0.2020	0.1335
nantes	1.2502	0.9106	1.6383	1.0843	0.9961	1.1797	0.2926	-0.0766	1.5830
nice	1.1255	0.8554	1.4127	1.0091	0.9980	1.0314	0.0570	-0.5099	0.9636
paris	1.7815	1.3652	2.2762	1.0524	1.0240	1.0879	0.1085	0.0538	0.1729
rennes	1.4659	0.9452	2.1707	1.0306	0.9395	1.1293	0.0896	-0.3871	0.9082
rouen	1.3011	0.9760	1.6338	1.0783	1.0131	1.1663	0.2576	0.0043	0.9911
strasbourg	1.5298	1.2436	1.8193	0.9850	0.9377	1.0251	-0.0449	-0.2148	0.0740
toulouse	1.2398	1.0020	1.5241	1.0525	1.0171	1.0983	0.2181	0.0574	0.7908

Results For Binary Ozone:

city	NDE	NDE_2.5	NDE_97.5	NIE	NIE_2.5	NIE_97.5	PMM	PMM_2.5	PMM_97.5
bordeaux	1.4841	1.2752	1.7013	8.8887	2.6331	41.8605	0.9600	0.8278	0.9930
clermont	1.4163	1.1848	1.6678	8.2661	1.9440	48.4141	0.9620	0.7678	0.9951
grenoble	1.1833	1.0378	1.3343	3.1608	0.6627	16.7460	0.9476	0.2158	1.5972
lehavre	1.3053	1.0667	1.5928	2.5796	0.2944	22.8162	0.9300	-1.1118	4.7234
lyon	1.4182	1.1801	1.6414	2.4633	0.8491	8.0006	0.8386	-0.6336	0.9721
marseille	1.1532	1.0565	1.2540	1.1816	0.9813	1.6420	0.5930	-0.1955	0.8699
montpellier	1.1002	0.9424	1.2836	1.1609	0.5908	2.3586	0.8209	-3.5830	4.3788
nancy	1.2322	1.0532	1.4399	1.1399	0.3048	4.2635	0.8406	-4.6805	6.2442
nantes	1.2492	1.0553	1.5183	24.6077	2.4769	373.8229	0.9925	0.8823	0.9996
nice	1.2149	1.0447	1.3933	1.4447	0.9278	2.6317	0.7252	-0.1852	0.9668
paris	1.7238	1.3804	2.1752	15.2836	6.0095	51.6234	0.9727	0.9283	0.9912
rennes	1.2685	1.0590	1.4783	37.6765	2.8967	579.5126	0.9943	0.8898	0.9998
rouen	1.3930	1.2034	1.6152	35.5309	4.0903	318.8118	0.9922	0.9125	0.9992
strasbourg	1.4272	1.2829	1.5698	7.7021	2.0842	38.4858	0.9580	0.7676	0.9935
toulouse	1.2554	1.1286	1.4090	3.3125	1.2356	10.4968	0.9204	0.5148	0.9816

Forestplot

Natural Direct Effect

Natural Indirect Effect

Analysis On Respiratory Mortality

Continuous Ozone

Code:

# some data management
 for (i in 1:length(villes_s)){
   villes_s[[i]]$time<-1:nrow(villes_s[[i]]) # Create a new variable for time 
 } 
   

    
  # start bootstrap
     
  nreps=800 #number of bootstrap reps
  results<-matrix(NA,nrow = 15,ncol=10)
  colnames(results)<-c("city","NDE","NDE_2.5","NDE_97.5","NIE","NIE_2.5","NIE_97.5","PMM","PMM_2.5","PMM_97.5")
  
for (i in (1:length(villes_s))){
  boot.nde<- rep(NA,nreps)
  boot.nie <- rep(NA,nreps)
  boot.pmm <- rep(NA,nreps)
  for (rep in 1:nreps)
      
    {
      dat_resample <- sample(1:nrow(villes_s[[i]]),nrow(villes_s[[i]]),replace=T)
      dat_boot <- villes_s[[i]][dat_resample,]
        
      # outcome model - Poisson Regression: Mortality, HW and O3
      m.outcome<-glm(nocc_tot~heat_wave+o3+ns(time,df=round(3*length(time)/122))+Jours+Vacances+hol,data=dat_boot,family=quasipoisson)
      
      # mediator model - multinomial linear regression
      m.mediator<-lm(o3~heat_wave+no2moy+ns(time,df=round(3*length(time)/122))+Jours+Vacances+hol,data=dat_boot) 
      
      # save coefficients
      theta1<-coef(m.outcome)[2]
      
      theta2<-coef(m.outcome)[3]
     
      beta1<-coef(m.mediator)[2]
      
      # estimate CDE and NIE
      
      boot.nde[rep] <- exp(theta1) #NDE
      
      boot.nie[rep] <-exp(theta2*beta1) #NIE 
      
      boot.pmm[rep] <-boot.nde[rep]*(boot.nie[rep]-1)/(boot.nde[rep]*boot.nie[rep]-1) #PMM 
                   
      
    } #end bootstrap

  results[i,1]<-cities[[i]]
  results[i,2]<-median(boot.nde, na.rm=T)
  results[i,3]<-quantile(boot.nde, 0.025, na.rm=T)
  results[i,4]<-quantile(boot.nde, 0.975, na.rm=T)
  results[i,5]<-median(boot.nie, na.rm=T)
  results[i,6]<-quantile(boot.nie, 0.025, na.rm=T)
  results[i,7]<-quantile(boot.nie, 0.975, na.rm=T)
  results[i,8]<-median(boot.pmm, na.rm=T)
  results[i,9]<-quantile(boot.pmm, 0.025, na.rm=T)
  results[i,10]<-quantile(boot.pmm, 0.975, na.rm=T)
       }
  
  
  # output CDE and NIE and confidence intervals

Binary Ozone

Code:

  # start bootstrap
     
  nreps=800 #number of bootstrap reps
  results_bin<-matrix(NA,nrow = 15,ncol=10)
  colnames(results_bin)<-c("city","NDE","NDE_2.5","NDE_97.5","NIE","NIE_2.5","NIE_97.5","PMM","PMM_2.5","PMM_97.5")
  
for (i in (1:length(villes_s))){
  boot.nde<- rep(NA,nreps)
  boot.nie <- rep(NA,nreps)
  boot.pmm <- rep(NA,nreps)
  for (rep in 1:nreps)
      
    {
      dat_resample <- sample(1:nrow(villes_s[[i]]),nrow(villes_s[[i]]),replace=T)
      dat_boot <- villes_s[[i]][dat_resample,] 
        
      # outcome model - Poisson Regression: Mortality, HW and O3
      m.outcome<-glm(nocc_tot~heat_wave+o3_90+ns(time,df=round(3*length(time)/122))+Jours+Vacances+hol,data=dat_boot,family=quasipoisson)
      
      # mediator model - multinomial linear regression
      m.mediator<-lm(o3~heat_wave+no2moy+ns(time,df=round(3*length(time)/122))+Jours+Vacances+hol,data=dat_boot) 
      
      # save coefficients
      theta1<-coef(m.outcome)[2]
      
      theta2<-coef(m.outcome)[3]
     
      beta1<-coef(m.mediator)[2]
      
      # estimate CDE and NIE
      
      boot.nde[rep] <- exp(theta1) #NDE
      
      boot.nie[rep] <-exp(theta2*beta1) #NIE 
      
      boot.pmm[rep] <-boot.nde[rep]*(boot.nie[rep]-1)/(boot.nde[rep]*boot.nie[rep]-1) #PMM 
                   
      
    } #end bootstrap

  results_bin[i,1]<-cities[[i]]
  results_bin[i,2]<-median(boot.nde, na.rm=T)
  results_bin[i,3]<-quantile(boot.nde, 0.025, na.rm=T)
  results_bin[i,4]<-quantile(boot.nde, 0.975, na.rm=T)
  results_bin[i,5]<-median(boot.nie, na.rm=T)
  results_bin[i,6]<-quantile(boot.nie, 0.025, na.rm=T)
  results_bin[i,7]<-quantile(boot.nie, 0.975, na.rm=T)
  results_bin[i,8]<-median(boot.pmm, na.rm=T)
  results_bin[i,9]<-quantile(boot.pmm, 0.025, na.rm=T)
  results_bin[i,10]<-quantile(boot.pmm, 0.975, na.rm=T)
       }
  
  
  # output CDE and NIE and confidence intervals

Results

Results For Continuous Ozone:

city	NDE	NDE_2.5	NDE_97.5	NIE	NIE_2.5	NIE_97.5	PMM	PMM_2.5	PMM_97.5
bordeaux	1.5050	1.2832	1.7276	1.0290	1.0048	1.0547	0.0793	0.0138	0.1613
clermont	1.4575	1.2424	1.7386	1.0182	0.9934	1.0492	0.0541	-0.0197	0.1569
grenoble	1.1607	1.0357	1.3005	1.0319	1.0109	1.0594	0.1811	0.0593	0.5154
lehavre	1.2826	1.0664	1.5591	1.0313	0.9845	1.0809	0.1272	-0.0703	0.4345
lyon	1.3855	1.1770	1.6093	1.0316	1.0165	1.0533	0.1039	0.0525	0.1865
marseille	1.1504	1.0623	1.2458	1.0044	0.9994	1.0120	0.0335	-0.0041	0.1109
montpellier	1.1056	0.9316	1.2813	1.0066	0.9946	1.0233	0.0522	-0.5118	0.8170
nancy	1.2491	1.0625	1.4763	1.0000	0.9790	1.0240	0.0005	-0.1528	0.1475
nantes	1.2142	1.0268	1.4716	1.0701	1.0315	1.1185	0.2900	0.1149	0.7626
nice	1.2177	1.0558	1.3937	1.0044	0.9987	1.0145	0.0249	-0.0069	0.1193
paris	1.7142	1.3965	2.1080	1.0527	1.0297	1.0819	0.1117	0.0688	0.1623
rennes	1.2778	1.0931	1.5216	1.0548	1.0084	1.1068	0.2050	0.0320	0.4856
rouen	1.3878	1.1961	1.5856	1.0677	1.0327	1.1092	0.1989	0.1024	0.3290
strasbourg	1.4572	1.3111	1.6261	1.0228	0.9999	1.0485	0.0673	-0.0004	0.1419
toulouse	1.2404	1.1066	1.3871	1.0333	1.0117	1.0572	0.1484	0.0586	0.2961

Results For Binary Ozone:

city	NDE	NDE_2.5	NDE_97.5	NIE	NIE_2.5	NIE_97.5	PMM	PMM_2.5	PMM_97.5
bordeaux	1.4884	1.2809	1.7147	9.5017	2.5775	42.7266	0.9641	0.8230	0.9925
clermont	1.4059	1.1814	1.6595	7.9246	1.7550	50.3589	0.9626	0.7004	0.9953
grenoble	1.1821	1.0369	1.3463	3.0242	0.6917	12.8590	0.9410	0.0130	1.6424
lehavre	1.2952	1.0518	1.5673	2.6227	0.2832	26.3602	0.9352	-2.5239	3.3444
lyon	1.4067	1.2263	1.6264	2.3063	0.8004	7.2868	0.8334	-0.3017	0.9719
marseille	1.1503	1.0437	1.2595	1.1967	0.9620	1.7491	0.6133	-0.3138	0.8829
montpellier	1.1137	0.9498	1.2964	1.1958	0.5976	2.3984	0.8153	-3.8212	3.7594
nancy	1.2364	1.0645	1.4529	1.1312	0.3020	4.5906	0.8364	-3.6178	5.9662
nantes	1.2397	1.0337	1.4809	24.5893	2.8541	293.7958	0.9927	0.9064	0.9997
nice	1.2053	1.0432	1.3894	1.4174	0.9059	2.8411	0.7220	-0.6344	0.9785
paris	1.7291	1.3905	2.1342	15.7180	6.4250	55.4055	0.9731	0.9300	0.9920
rennes	1.2613	1.0730	1.4972	33.9426	3.0054	633.0013	0.9940	0.8957	0.9998
rouen	1.3948	1.2153	1.5980	34.9744	4.3325	314.7414	0.9921	0.9205	0.9992
strasbourg	1.4303	1.2697	1.5754	7.7439	2.0084	38.0677	0.9577	0.7610	0.9935
toulouse	1.2528	1.1139	1.3932	3.2039	1.1726	10.6449	0.9185	0.4458	0.9841

Forestplot

Natural Direct Effect

Natural Indirect Effect

Heat wave and Ozone, Mediation analysis on mortality (Meteo France), test for non linearity

Anna Alari

24/11/2020

Temperature Data Management

Creating Database only with summer months

Creating Heat Wave variable

Creating Ozone binary variable

Analysis On Non-Accidental Mortality

Continuous Ozone

Binary Ozone

Results

Forestplot

Analysis On Cardiovascual Mortality

Continuous Ozone

Binary Ozone

Results

Forestplot

Analysis On Respiratory Mortality

Continuous Ozone

Binary Ozone

Results

Forestplot