LMLS 2020-12-24
library(tidyverse)## ── Attaching packages ─────────────────────────────────────────────────────────────────────────────────────────────────────────────────────── tidyverse 1.3.0 ──
## ✓ ggplot2 3.3.2 ✓ purrr 0.3.4
## ✓ tibble 3.0.3 ✓ dplyr 1.0.2
## ✓ tidyr 1.1.2 ✓ stringr 1.4.0
## ✓ readr 1.3.1 ✓ forcats 0.5.0
## ── Conflicts ────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────── tidyverse_conflicts() ──
## x dplyr::filter() masks stats::filter()
## x dplyr::lag() masks stats::lag()
library(caret)## Loading required package: lattice
##
## Attaching package: 'caret'
## The following object is masked from 'package:purrr':
##
## lift
setwd("~/Documentos/R/Stroke/death/")
y2016deaths <- read.csv("yearsdeaths.cvs")
glimpse(y2016deaths)## Rows: 6,566
## Columns: 9
## $ sexo <chr> "Women", "Women", "Men", "Men", "Women", "Women", "Men", "M…
## $ edad <int> 70, 87, 80, 66, 57, 56, 64, 64, 37, 90, 73, 98, 85, 92, 92,…
## $ cid10 <chr> "I64", "I64", "I64", "I64", "I64", "I64", "I60", "I60", "I6…
## $ agecat <chr> "55-74", "75+", "75+", "55-74", "55-74", "55-74", "55-74", …
## $ level <chr> "I, II", "I, II", "III", "III", "III", "III", "III", "III",…
## $ regions <chr> "Sierra", "Sierra", "Sierra", "Sierra", "Sierra", "Sierra",…
## $ condicion <chr> "Discharged", "Discharged", "Discharged", "Discharged", "Di…
## $ los <int> 2, 1, 28, 14, 2, 7, 10, 3, 12, 11, 12, 5, 4, 5, 12, 8, 19, …
## $ anio <int> 2017, 2017, 2017, 2017, 2017, 2017, 2017, 2017, 2017, 2017,…
y2016deaths$exposure <- ifelse(y2016deaths$cid10=="I63",0,
ifelse(y2016deaths$cid10=="I60",1,
ifelse(y2016deaths$cid10=="I61",2,
ifelse(y2016deaths$cid10=="I64",3,NA))))
y2016deaths$exposure <- factor(y2016deaths$exposure, levels = c(0,1,2,3), labels = c("I63", "I60", "I61","I64"))
sum(is.na(y2016deaths$exposure)) #0## [1] 0
y2016deaths$outcome <- ifelse(y2016deaths$condicion=="Discharged", 0,
ifelse(y2016deaths$condicion=="Death", 1, NA)) %>%
factor(y2016deaths$outcome, levels = c(0,1), labels = c("Discharged","Death"))glm.fit <- glm(outcome ~ exposure + sexo + edad,
data = y2016deaths,
family = binomial)glm.fit.1 <- glm(outcome ~ exposure + sexo + edad +level,
data = y2016deaths,
family = binomial, subset = y2016deaths$regions=="Lima/Callao")
glm.fit.2 <- glm(outcome ~ exposure + sexo + edad +level,
data = y2016deaths,
family = binomial, subset = y2016deaths$regions=="Resto Costa")
glm.fit.3 <- glm(outcome ~ exposure + sexo + edad +level,
data = y2016deaths,
family = binomial, subset = y2016deaths$regions=="Selva")
glm.fit.4 <- glm(outcome ~ exposure + sexo + edad +level,
data = y2016deaths,
family = binomial, subset = y2016deaths$regions=="Sierra")
glm.fit.5 <- glm(outcome ~ exposure + sexo + edad +regions,
data = y2016deaths,
family = binomial, subset = y2016deaths$level=="I, II")
glm.fit.6 <- glm(outcome ~ exposure + sexo + edad +regions,
data = y2016deaths,
family = binomial, subset = y2016deaths$level=="III")summary(glm.fit)##
## Call:
## glm(formula = outcome ~ exposure + sexo + edad, family = binomial,
## data = y2016deaths)
##
## Deviance Residuals:
## Min 1Q Median 3Q Max
## -0.8725 -0.5266 -0.4483 -0.3729 2.5386
##
## Coefficients:
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) -3.756189 0.232665 -16.144 < 2e-16 ***
## exposureI60 1.399018 0.136506 10.249 < 2e-16 ***
## exposureI61 0.904375 0.122736 7.368 1.73e-13 ***
## exposureI64 0.414292 0.112511 3.682 0.000231 ***
## sexoWomen 0.196411 0.079564 2.469 0.013565 *
## edad 0.015121 0.002984 5.067 4.05e-07 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for binomial family taken to be 1)
##
## Null deviance: 4693.9 on 6565 degrees of freedom
## Residual deviance: 4545.1 on 6560 degrees of freedom
## AIC: 4557.1
##
## Number of Fisher Scoring iterations: 5
exp(coef(glm.fit))## (Intercept) exposureI60 exposureI61 exposureI64 sexoWomen edad
## 0.02337265 4.05122049 2.47038813 1.51329933 1.21702746 1.01523582
exp(confint(glm.fit))## Waiting for profiling to be done...
## 2.5 % 97.5 %
## (Intercept) 0.01474398 0.03671004
## exposureI60 3.10233290 5.29990044
## exposureI61 1.94521515 3.14845240
## exposureI64 1.21700336 1.89232339
## sexoWomen 1.04155262 1.42291031
## edad 1.00934399 1.02122358
##model1
summary(glm.fit.1)##
## Call:
## glm(formula = outcome ~ exposure + sexo + edad + level, family = binomial,
## data = y2016deaths, subset = y2016deaths$regions == "Lima/Callao")
##
## Deviance Residuals:
## Min 1Q Median 3Q Max
## -1.0885 -0.4811 -0.3881 -0.3197 2.5495
##
## Coefficients:
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) -3.06811 0.37939 -8.087 6.12e-16 ***
## exposureI60 1.49631 0.18554 8.065 7.34e-16 ***
## exposureI61 0.68811 0.17725 3.882 0.000104 ***
## exposureI64 0.13869 0.17361 0.799 0.424378
## sexoWomen 0.37372 0.12612 2.963 0.003045 **
## edad 0.01247 0.00468 2.665 0.007706 **
## levelIII -0.77833 0.15753 -4.941 7.78e-07 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for binomial family taken to be 1)
##
## Null deviance: 2002.5 on 3079 degrees of freedom
## Residual deviance: 1895.6 on 3073 degrees of freedom
## AIC: 1909.6
##
## Number of Fisher Scoring iterations: 5
exp(coef(glm.fit.1))## (Intercept) exposureI60 exposureI61 exposureI64 sexoWomen edad
## 0.04650882 4.46516642 1.98994441 1.14876723 1.45313045 1.01254956
## levelIII
## 0.45917471
exp(confint(glm.fit.1))## Waiting for profiling to be done...
## 2.5 % 97.5 %
## (Intercept) 0.02191171 0.09701797
## exposureI60 3.10784282 6.43830128
## exposureI61 1.40631512 2.82088118
## exposureI64 0.81799040 1.61742529
## sexoWomen 1.13611486 1.86346300
## edad 1.00335038 1.02193788
## levelIII 0.33855148 0.62832664
##model2
summary(glm.fit.2)##
## Call:
## glm(formula = outcome ~ exposure + sexo + edad + level, family = binomial,
## data = y2016deaths, subset = y2016deaths$regions == "Resto Costa")
##
## Deviance Residuals:
## Min 1Q Median 3Q Max
## -0.8209 -0.5407 -0.4713 -0.3166 2.6605
##
## Coefficients:
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) -4.061311 0.491422 -8.264 < 2e-16 ***
## exposureI60 1.488841 0.352374 4.225 2.39e-05 ***
## exposureI61 1.507906 0.270070 5.583 2.36e-08 ***
## exposureI64 0.921608 0.281522 3.274 0.00106 **
## sexoWomen 0.101671 0.153344 0.663 0.50731
## edad 0.014231 0.005602 2.540 0.01108 *
## levelIII -0.074570 0.181259 -0.411 0.68078
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for binomial family taken to be 1)
##
## Null deviance: 1267.0 on 1794 degrees of freedom
## Residual deviance: 1221.6 on 1788 degrees of freedom
## AIC: 1235.6
##
## Number of Fisher Scoring iterations: 5
exp(coef(glm.fit.2))## (Intercept) exposureI60 exposureI61 exposureI64 sexoWomen edad
## 0.01722641 4.43195586 4.51726010 2.51332884 1.10701948 1.01433277
## levelIII
## 0.92814221
exp(confint(glm.fit.2))## Waiting for profiling to be done...
## 2.5 % 97.5 %
## (Intercept) 0.006432556 0.04423645
## exposureI60 2.213772212 8.88712262
## exposureI61 2.714712281 7.86952121
## exposureI64 1.476531720 4.47325230
## sexoWomen 0.820026137 1.49694004
## edad 1.003360301 1.02565758
## levelIII 0.649971975 1.32346968
##model3
summary(glm.fit.3)##
## Call:
## glm(formula = outcome ~ exposure + sexo + edad + level, family = binomial,
## data = y2016deaths, subset = y2016deaths$regions == "Selva")
##
## Deviance Residuals:
## Min 1Q Median 3Q Max
## -1.2181 -0.4502 -0.4044 -0.3908 2.3306
##
## Coefficients:
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) -2.500264 0.824933 -3.031 0.002439 **
## exposureI60 0.529110 0.488843 1.082 0.279088
## exposureI61 0.657892 0.489440 1.344 0.178893
## exposureI64 -0.273555 0.397621 -0.688 0.491465
## sexoWomen -0.037676 0.280147 -0.134 0.893017
## edad 0.004208 0.010860 0.387 0.698392
## levelIII 1.672433 0.435920 3.837 0.000125 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for binomial family taken to be 1)
##
## Null deviance: 411.07 on 579 degrees of freedom
## Residual deviance: 380.18 on 573 degrees of freedom
## (1 observation deleted due to missingness)
## AIC: 394.18
##
## Number of Fisher Scoring iterations: 5
exp(coef(glm.fit.3))## (Intercept) exposureI60 exposureI61 exposureI64 sexoWomen edad
## 0.0820633 1.6974201 1.9307184 0.7606702 0.9630247 1.0042171
## levelIII
## 5.3251087
exp(confint(glm.fit.3))## Waiting for profiling to be done...
## 2.5 % 97.5 %
## (Intercept) 0.01533718 0.3928546
## exposureI60 0.65069355 4.4994162
## exposureI61 0.73568364 5.1140594
## exposureI64 0.36001247 1.7363666
## sexoWomen 0.55274629 1.6648286
## edad 0.98332553 1.0262139
## levelIII 2.24913599 12.5643336
##model4
summary(glm.fit.4)##
## Call:
## glm(formula = outcome ~ exposure + sexo + edad + level, family = binomial,
## data = y2016deaths, subset = y2016deaths$regions == "Sierra")
##
## Deviance Residuals:
## Min 1Q Median 3Q Max
## -1.0023 -0.6138 -0.5520 -0.4551 2.3853
##
## Coefficients:
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) -3.216433 0.539194 -5.965 2.44e-09 ***
## exposureI60 1.014355 0.311933 3.252 0.00115 **
## exposureI61 0.641726 0.291842 2.199 0.02789 *
## exposureI64 -0.004345 0.245606 -0.018 0.98589
## sexoWomen 0.093928 0.166205 0.565 0.57199
## edad 0.018889 0.006553 2.882 0.00395 **
## levelIII -0.338013 0.223875 -1.510 0.13109
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for binomial family taken to be 1)
##
## Null deviance: 983.99 on 1109 degrees of freedom
## Residual deviance: 957.17 on 1103 degrees of freedom
## AIC: 971.17
##
## Number of Fisher Scoring iterations: 4
exp(coef(glm.fit.4))## (Intercept) exposureI60 exposureI61 exposureI64 sexoWomen edad
## 0.04009782 2.75758294 1.89975691 0.99566476 1.09848025 1.01906845
## levelIII
## 0.71318609
exp(confint(glm.fit.4))## Waiting for profiling to be done...
## 2.5 % 97.5 %
## (Intercept) 0.01358141 0.1126879
## exposureI60 1.49954225 5.1162951
## exposureI61 1.07673245 3.3946731
## exposureI64 0.62356597 1.6391252
## sexoWomen 0.79355754 1.5236787
## edad 1.00626728 1.0324802
## levelIII 0.45286056 1.0920554
##model5
summary(glm.fit.5)##
## Call:
## glm(formula = outcome ~ exposure + sexo + edad + regions, family = binomial,
## data = y2016deaths, subset = y2016deaths$level == "I, II")
##
## Deviance Residuals:
## Min 1Q Median 3Q Max
## -0.9969 -0.5697 -0.4851 -0.4063 2.5269
##
## Coefficients:
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) -3.239249 0.361577 -8.959 < 2e-16 ***
## exposureI60 1.097073 0.236406 4.641 3.47e-06 ***
## exposureI61 0.753231 0.202442 3.721 0.000199 ***
## exposureI64 0.101723 0.171856 0.592 0.553912
## sexoWomen 0.144382 0.112032 1.289 0.197482
## edad 0.017090 0.004273 4.000 6.34e-05 ***
## regionsResto Costa -0.327105 0.167736 -1.950 0.051162 .
## regionsSelva -0.560717 0.199897 -2.805 0.005031 **
## regionsSierra 0.036304 0.162405 0.224 0.823117
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for binomial family taken to be 1)
##
## Null deviance: 2278.2 on 2884 degrees of freedom
## Residual deviance: 2203.9 on 2876 degrees of freedom
## (1 observation deleted due to missingness)
## AIC: 2221.9
##
## Number of Fisher Scoring iterations: 5
exp(coef(glm.fit.5))## (Intercept) exposureI60 exposureI61 exposureI64
## 0.03919332 2.99538671 2.12385180 1.10707647
## sexoWomen edad regionsResto Costa regionsSelva
## 1.15532511 1.01723636 0.72100800 0.57079970
## regionsSierra
## 1.03697075
exp(confint(glm.fit.5))## Waiting for profiling to be done...
## 2.5 % 97.5 %
## (Intercept) 0.01908401 0.07879767
## exposureI60 1.88313535 4.76663145
## exposureI61 1.43327861 3.17430391
## exposureI64 0.79703533 1.56541478
## sexoWomen 0.92785134 1.43986278
## edad 1.00882807 1.02587453
## regionsResto Costa 0.52040948 1.00530464
## regionsSelva 0.38446791 0.84305342
## regionsSierra 0.75691814 1.43185806
##model6
summary(glm.fit.6)##
## Call:
## glm(formula = outcome ~ exposure + sexo + edad + regions, family = binomial,
## data = y2016deaths, subset = y2016deaths$level == "III")
##
## Deviance Residuals:
## Min 1Q Median 3Q Max
## -1.3421 -0.4955 -0.3878 -0.3239 2.5482
##
## Coefficients:
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) -3.722895 0.323951 -11.492 < 2e-16 ***
## exposureI60 1.449754 0.173116 8.374 < 2e-16 ***
## exposureI61 0.938787 0.157055 5.977 2.27e-09 ***
## exposureI64 0.363652 0.170518 2.133 0.0330 *
## sexoWomen 0.245344 0.114570 2.141 0.0322 *
## edad 0.010115 0.004271 2.368 0.0179 *
## regionsResto Costa 0.167030 0.137909 1.211 0.2258
## regionsSelva 1.709119 0.384479 4.445 8.78e-06 ***
## regionsSierra 0.384566 0.212570 1.809 0.0704 .
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for binomial family taken to be 1)
##
## Null deviance: 2397.0 on 3679 degrees of freedom
## Residual deviance: 2271.8 on 3671 degrees of freedom
## (1 observation deleted due to missingness)
## AIC: 2289.8
##
## Number of Fisher Scoring iterations: 5
exp(coef(glm.fit.6))## (Intercept) exposureI60 exposureI61 exposureI64
## 0.0241639 4.2620666 2.5568782 1.4385733
## sexoWomen edad regionsResto Costa regionsSelva
## 1.2780608 1.0101668 1.1817896 5.5240950
## regionsSierra
## 1.4689769
exp(confint(glm.fit.6))## Waiting for profiling to be done...
## 2.5 % 97.5 %
## (Intercept) 0.01269958 0.04523753
## exposureI60 3.04003454 5.99765873
## exposureI61 1.88402601 3.49023777
## exposureI64 1.02963935 2.01136320
## sexoWomen 1.02161260 1.60129541
## edad 1.00178443 1.01870625
## regionsResto Costa 0.89800677 1.54283467
## regionsSelva 2.56592316 11.72838160
## regionsSierra 0.95249752 2.19742849