The goal of decompositionLE is to provide an easy to use implementation of life expectancy decomposition formulas for age bands, derived from Ponnapalli, K. (2005) In addition, there is a decomposition function for disease cause breakdown sourced from Preston, S.H., Heuveline, P. and Guillot, M. (2001), as well as some useful helper plot functions.
You can install the development version of decompositionLE from GitHub with:
# devtools::install_github("herts-phei/decompositionLE")
install.packages("decompositionLE", repos = c('https://herts-phei.r-universe.dev', 'https://cloud.r-project.org'))
The package contains a built-in dataset of life table values for US
women born in 1935 and 1995, us_females, sourced from Ponnapalli, K.
(2005).
us_females
#> Age nm1x l1x e1x nm2x l2x e2x
#> 1 0 0.047139 1.00000 63.32 0.006830 1.00000 79.00
#> 2 1 0.004157 0.95458 65.32 0.000358 0.99321 78.54
#> 3 5 0.001525 0.93887 62.39 0.000167 0.99179 74.65
#> 4 10 0.001208 0.93174 57.85 0.000196 0.99096 69.71
#> 5 15 0.002022 0.92613 53.19 0.000459 0.98999 64.78
#> 6 20 0.002944 0.91681 48.70 0.000503 0.98772 59.92
#> 7 25 0.003562 0.90341 44.38 0.000647 0.98524 55.06
#> 8 30 0.003980 0.88746 40.14 0.000892 0.98206 50.23
#> 9 35 0.005003 0.86997 35.89 0.001266 0.97769 45.45
#> 10 40 0.005927 0.84847 31.74 0.001765 0.97152 40.72
#> 11 45 0.008310 0.82368 27.61 0.002612 0.96298 36.06
#> 12 50 0.011638 0.79012 23.67 0.004171 0.95048 31.49
#> 13 55 0.016309 0.74539 19.94 0.006575 0.93085 27.10
#> 14 60 0.024373 0.68688 16.41 0.010387 0.90071 22.92
#> 15 65 0.035823 0.60779 13.21 0.016349 0.85504 19.00
#> 16 70 0.055769 0.50757 10.31 0.024504 0.78775 15.40
#> 17 75 0.092454 0.38276 7.82 0.038841 0.69655 12.07
#> 18 80 0.130808 0.23930 6.03 0.063900 0.57275 9.12
#> 19 85+ 0.222482 0.12281 4.49 0.151106 0.41424 6.62
The dataset demonstrates usage of decomp_age().
age_output <- decomp_age(us_females, method = "arriaga3",
age_col = "Age", e1 = "e1x", e2 = "e2x", l1 = "l1x", l2 = "l2x")
age_output
#> Age nm1x l1x e1x nm2x l2x e2x direct_effect
#> 1 0 0.047139 1.00000 63.32 0.006830 1.00000 79.00 0.02645220
#> 2 1 0.004157 0.95458 65.32 0.000358 0.99321 78.54 0.03890153
#> 3 5 0.001525 0.93887 62.39 0.000167 0.99179 74.65 0.01835259
#> 4 10 0.001208 0.93174 57.85 0.000196 0.99096 69.71 0.01264484
#> 5 15 0.002022 0.92613 53.19 0.000459 0.98999 64.78 0.01658115
#> 6 20 0.002944 0.91681 48.70 0.000503 0.98772 59.92 0.02818036
#> 7 25 0.003562 0.90341 44.38 0.000647 0.98524 55.06 0.04102099
#> 8 30 0.003980 0.88746 40.14 0.000892 0.98206 50.23 0.02330090
#> 9 35 0.005003 0.86997 35.89 0.001266 0.97769 45.45 0.04856497
#> 10 40 0.005927 0.84847 31.74 0.001765 0.97152 40.72 0.03666408
#> 11 45 0.008310 0.82368 27.61 0.002612 0.96298 36.06 0.06641661
#> 12 50 0.011638 0.79012 23.67 0.004171 0.95048 31.49 0.07906885
#> 13 55 0.016309 0.74539 19.94 0.006575 0.93085 27.10 0.08717324
#> 14 60 0.024373 0.68688 16.41 0.010387 0.90071 22.92 0.12886018
#> 15 65 0.035823 0.60779 13.21 0.016349 0.85504 19.00 0.15499332
#> 16 70 0.055769 0.50757 10.31 0.024504 0.78775 15.40 0.20366438
#> 17 75 0.092454 0.38276 7.82 0.038841 0.69655 12.07 0.28108338
#> 18 80 0.130808 0.23930 6.03 0.063900 0.57275 9.12 0.24621067
#> 19 85+ 0.222482 0.12281 4.49 0.151106 0.41424 6.62 0.57195825
#> indirect_effect exclusive_effect interaction_effect total_effect
#> 1 2.7786559 2.8051081 0.000000000 2.8051081
#> 2 1.0046032 1.0435047 -0.003558610 1.0399461
#> 3 0.4171028 0.4354554 -0.002120567 0.4333348
#> 4 0.2867811 0.2994260 -0.001730363 0.2976956
#> 5 0.4057047 0.4222859 -0.002783792 0.4195021
#> 6 0.5754256 0.6036060 -0.004583696 0.5990223
#> 7 0.6185978 0.6596188 -0.005956216 0.6536626
#> 8 0.5835122 0.6068131 -0.006899530 0.5999136
#> 9 0.6203954 0.6689604 -0.008900702 0.6600597
#> 10 0.5970722 0.6337363 -0.010619689 0.6231166
#> 11 0.6915976 0.7580142 -0.015121718 0.7428925
#> 12 0.7463816 0.8254504 -0.020643620 0.8048068
#> 13 0.7740000 0.8611732 -0.027839442 0.8333338
#> 14 0.8437428 0.9726030 -0.039890424 0.9327126
#> 15 0.8375481 0.9925414 -0.054237938 0.9383034
#> 16 0.8768650 1.0805294 -0.077473759 1.0030556
#> 17 0.8533742 1.1344576 -0.095541086 1.0389165
#> 18 0.5110394 0.7572501 -0.074591117 0.6826590
#> 19 NA 0.5719583 NA 0.5719583
In decomp_age(), there are several methods currently available:
arriaga3, chandrasekaran1 or chandrasekaran2. Functionally, they
produce close to identical results, but the output variables are
different.
age_col should be a factor column providing ordered age bands with the
final age group being an open-ended interval suffixed with ‘+’,
e.g. 90+. e1 and l1 correspond to the expectation of life at age
group x, and the proportion of persons alive at age group x in decimal
form, in the first group of comparison, and likewise for e2 and l2
for the second group of comparison.
There is also a helpful plot function for output from decomp_age()
offering a variety of basic ggplot2 style plots.
library(patchwork)
plot1 <- plot_age(age_output, method = "arriaga3", plot_type = "segment_dodge")
plot2 <- plot_age(age_output, method = "arriaga3", plot_type = "segment_dodge", line = TRUE)
plot3 <- plot_age(age_output, method = "arriaga3", plot_type = "total", line = TRUE)
plot1
plot2
plot3
The package also contains a built-in dataset of age and cause
decomposition of difference in Life Expectancies at birth, for India and
China, males, 1990, india_china_males_1990, sourced from Murray,
C.J.L. and Lopez, A.D. (1996).
india_china_males_1990
#> Age India_nmx India_CD India_NCD India_Injuries China_nmx China_CD China_NCD
#> 1 0 0.0267 0.882 0.073 0.046 0.0084 0.677 0.174
#> 2 5 0.0025 0.504 0.188 0.309 0.0009 0.174 0.337
#> 3 15 0.0021 0.382 0.223 0.394 0.0015 0.068 0.380
#> 4 30 0.0043 0.429 0.315 0.257 0.0028 0.101 0.573
#> 5 45 0.0139 0.304 0.592 0.104 0.0102 0.095 0.796
#> 6 60 0.0388 0.248 0.722 0.030 0.0342 0.070 0.879
#> 7 70+ 0.0929 0.247 0.728 0.025 0.1003 0.084 0.877
#> China_Injuries nDx
#> 1 0.149 5.6
#> 2 0.488 0.8
#> 3 0.552 0.3
#> 4 0.326 0.6
#> 5 0.109 0.8
#> 6 0.051 0.3
#> 7 0.039 -0.3
The dataset demonstrates usage of decomp_disease().
disease_output <- decomp_disease(india_china_males_1990,
breakdown = "proportion",
age_col = "Age",
diseases = c("CD", "NCD", "Injuries"),
group_1 = "India", group_1_m = "India_nmx",
group_2 = "China", group_2_m = "China_nmx",
nDx = "nDx"
)
disease_output
#> Age India_nmx India_CD India_NCD India_Injuries China_nmx China_CD China_NCD
#> 1 0 0.0267 0.882 0.073 0.046 0.0084 0.677 0.174
#> 2 5 0.0025 0.504 0.188 0.309 0.0009 0.174 0.337
#> 3 15 0.0021 0.382 0.223 0.394 0.0015 0.068 0.380
#> 4 30 0.0043 0.429 0.315 0.257 0.0028 0.101 0.573
#> 5 45 0.0139 0.304 0.592 0.104 0.0102 0.095 0.796
#> 6 60 0.0388 0.248 0.722 0.030 0.0342 0.070 0.879
#> 7 70+ 0.0929 0.247 0.728 0.025 0.1003 0.084 0.877
#> China_Injuries nDx delta_CD delta_NCD delta_Injuries
#> 1 0.149 5.6 5.4661508 0.1491803 -0.007160656
#> 2 0.488 0.8 0.5517000 0.0833500 0.166650000
#> 3 0.552 0.3 0.3501000 -0.0508500 -0.000300000
#> 4 0.326 0.6 0.6247600 -0.0999600 0.076920000
#> 5 0.109 0.8 0.7041297 0.0236973 0.072172973
#> 6 0.051 0.3 0.4714174 -0.1335783 -0.037839130
#> 7 0.039 -0.3 0.5886932 -0.8242662 -0.064427027
The diseases argument provides the function with suffixes of relevant
diseases found in both groups of interest. breakdown is for whether
disease breakdowns are raw mortality rates or a decimal proportion
of the total all-cause mortality rate. For example, in the
india_china_males_1990 dataset which has decimal proportion
breakdowns:
india_china_males_1990[1,]
#> Age India_nmx India_CD India_NCD India_Injuries China_nmx China_CD China_NCD
#> 1 0 0.0267 0.882 0.073 0.046 0.0084 0.677 0.174
#> China_Injuries nDx
#> 1 0.149 5.6
At birth, the components which make up India’s all-cause mortality in the dataset:
india_birth_components_equal_1 <- india_china_males_1990[1,c("India_CD", "India_NCD", "India_Injuries")]
all.equal(sum(india_birth_components_equal_1), 1, tolerance = 1e-3)
#> [1] TRUE
Within tolerance, is equal to 1. Likewise, for China all-cause mortality at birth:
china_birth_components_equal_1 <- india_china_males_1990[1,c("China_CD", "China_NCD", "China_Injuries")]
all.equal(sum(china_birth_components_equal_1), 1, tolerance = 1e-3)
#> [1] TRUE
However raw disease breakdowns are also accepted if provided, but these disease breakdowns should then equal the value provided in the column for the all-cause mortality rate for the specified group instead, not 1.
Likewise in decomp_age(), age_col should be a factor column
providing ordered age bands with the final age group being an open-ended
interval suffixed with ‘+’.
The diseases argument specifies a character vector of the diseases
found in both groups, and should be suffixed at the end of the names
given in group_1 and group_2 in the dataset. In
india_china_males_1990, these are: c("CD", "NCD", "Injuries").
There is a also a helper plot function for disease decomposition output,
plot_disease():
plot_disease(x = disease_output, suffixes = c("delta_CD", "delta_NCD", "delta_Injuries"), nDx = "nDx", line = FALSE)
plot_disease(x = disease_output, suffixes = c("delta_CD", "delta_NCD", "delta_Injuries"), nDx = "nDx", line = TRUE)
Ponnapalli, K. (2005). A comparison of different methods for decomposition of changes in expectation of life at birth and differentials in life expectancy at birth. Demographic Research, 12, pp.141–172. doi: https://doi.org/10.4054/demres.2005.12.7.
Preston, S.H., Heuveline, P. and Guillot, M. (2001). Demography : measuring and modeling population processes. Oxford ; Malden, Mass.: Blackwell Publishers.