2020 excess mortality & voting patterns in CH

Assesing bias & coverage of 2015-19 models

Data

Mortality data

data = read_rds("data/BfS-closed/monthly_deaths/w_deaths_2015_2020_year_fin.Rds") %>% 
  filter(age != "<40") %>% 
  # strata with double zeroes seem to be crashing
  filter(pop_mid_poi > 0)  %>% 
  # only data till 2019
  filter(year < 2020)

data_male = data %>% 
  filter(sex == "Male") %>% 
  select(-sex) %>% 
  as.data.frame()

# data_female = data %>% 
#   filter(sex == "Female") %>% 
#   select(-sex) %>% 
#   as.data.frame()

Spatial

kt = read_rds("data/BfS/kt.Rds")
gg = read_rds("data/BfS/gg.Rds")
tg3o = read_rds("data/BfS/tg3o.Rds")
se_alt = read_rds("data/BfS/se_alt.Rds")

Results from Ubelix

gem_bym2_cv = read_rds("results_inla/gem_bym2_cv.Rds")
gem_bym2_cv_kt = read_rds("results_inla/gem_bym2_cv_kt.Rds")

Year of interest

yoi = 2019

Canton

Models with and without canton (iid).

DIC

WAIC

Models

No canton

summary(gem_bym2_cv$Male)


Call:
   c("inla(formula = formula, family = \"Poisson\", data = data_sex, ", " 
   quantiles = c(0.025, 0.5, 0.975), control.compute = list(config = TRUE, 
   ", " cpo = TRUE, dic = TRUE, waic = TRUE), control.predictor = 
   list(compute = TRUE, ", " link = 1), control.family = control.family, 
   control.mode = list(restart = TRUE), ", " num.threads = threads)") 
Time used:
    Pre = 13.6, Running = 122, Post = 5.17, Total = 141 
Fixed effects:
              mean    sd 0.025quant 0.5quant 0.975quant   mode kld
(Intercept) 47.676 5.441     37.000   47.673     58.357 47.669   0
year        -0.026 0.003     -0.031   -0.026     -0.021 -0.026   0

Random effects:
  Name    Model
    id_age IID model
   id_space BYM2 model
   row_id IID model

Model hyperparameters:
                          mean     sd 0.025quant 0.5quant 0.975quant    mode
Precision for id_age     0.816  0.343      0.357    0.747      1.672   0.630
Precision for id_space  56.842  5.074     47.279   56.729     67.178  56.638
Phi for id_space         0.394  0.075      0.267    0.387      0.557   0.364
Precision for row_id   255.840 64.190    154.199  247.545    404.124 231.688

Deviance Information Criterion (DIC) ...............: 109037.28
Deviance Information Criterion (DIC, saturated) ....: 40687.99
Effective number of parameters .....................: 1087.92

Watanabe-Akaike information criterion (WAIC) ...: 109055.96
Effective number of parameters .................: 1033.89

Marginal log-Likelihood:  -53882.05 
CPO, PIT is computed 
Posterior summaries for the linear predictor and the fitted values are computed
(Posterior marginals needs also 'control.compute=list(return.marginals.predictor=TRUE)')

Canton

summary(gem_bym2_cv_kt$Male)


Call:
   c("inla(formula = formula, family = \"Poisson\", data = data_sex, ", " 
   quantiles = c(0.025, 0.5, 0.975), control.compute = list(config = TRUE, 
   ", " cpo = TRUE, dic = TRUE, waic = TRUE), control.predictor = 
   list(compute = TRUE, ", " link = 1), control.family = control.family, 
   control.mode = list(restart = TRUE), ", " num.threads = threads)") 
Time used:
    Pre = 21.4, Running = 110, Post = 5.37, Total = 137 
Fixed effects:
              mean    sd 0.025quant 0.5quant 0.975quant   mode kld
(Intercept) 47.595 5.427     36.947   47.593     58.248 47.588   0
year        -0.026 0.003     -0.031   -0.026     -0.021 -0.026   0

Random effects:
  Name    Model
    id_age IID model
   id_kt IID model
   id_space BYM2 model
   row_id IID model

Model hyperparameters:
                           mean       sd 0.025quant 0.5quant 0.975quant    mode
Precision for id_age      0.646    0.219      0.303    0.619       1.15   0.564
Precision for id_kt    3199.311 5766.719    312.357 1610.714   16044.72 687.504
Precision for id_space   58.014    5.310     48.677   57.618      69.53  56.655
Phi for id_space          0.374    0.082      0.230    0.368       0.55   0.352
Precision for row_id    272.344   77.631    147.143  263.570     449.53 246.735

Deviance Information Criterion (DIC) ...............: 109036.78
Deviance Information Criterion (DIC, saturated) ....: 40687.49
Effective number of parameters .....................: 1062.98

Watanabe-Akaike information criterion (WAIC) ...: 109056.47
Effective number of parameters .................: 1012.88

Marginal log-Likelihood:  -53881.42 
CPO, PIT is computed 
Posterior summaries for the linear predictor and the fitted values are computed
(Posterior marginals needs also 'control.compute=list(return.marginals.predictor=TRUE)')

Canton effect

Observed vs predicted

No canton

Canton

Age bias

No canton

Canton

Coverage

No canton

frq(males_fit_2000_space$within)

x <character> 
# total N=10705 valid N=10705 mean=1.65 sd=0.48

Value |    N | Raw % | Valid % | Cum. %
---------------------------------------
no    | 3752 | 35.05 |   35.05 |  35.05
yes   | 6953 | 64.95 |   64.95 | 100.00
<NA>  |    0 |  0.00 |    <NA> |   <NA>

Canton

frq(males_fit_kt_2000_space$within)

x <character> 
# total N=10705 valid N=10705 mean=1.65 sd=0.48

Value |    N | Raw % | Valid % | Cum. %
---------------------------------------
no    | 3782 | 35.33 |   35.33 |  35.33
yes   | 6923 | 64.67 |   64.67 | 100.00
<NA>  |    0 |  0.00 |    <NA> |   <NA>