I run a multiple regression to examine the relationship between countries’ mortality rate and border closure policy type due to COVID-19.
Outcome: Countries’ mortarity rate.
Predictor: Border closure policy type during the pandemic year (complete closure vs. partial closure).
Controlled variables: test number, case number, critical case rate, government effectiveness score, proportion of population aged 65 or older, number of hospital beds, number of deaths attributable to communicable diseases, and transport infrastructure quality score.
# Source: https://doi.org/10.7910/DVN/U6DJAC updated May 2, 2021
cobap <- rio::import(here::here("data", "policy_list.csv")) %>% select(ISO3, COUNTRY_NAME, POLICY_TYPE, POLICY_SUBTYPE, START_DATE, END_DATE)
control_data <- readRDS("coviddata.RDS") # control Variables - This data is prepared by Mark A. Weiss
data <- left_join(cobap, control_data ) %>%
drop_na()
#convert some columns to correct types
data[, 9:ncol(data)] <- lapply(9:ncol(data), function(x)
as.numeric(data[[x]]))
data[, 3:4] <- lapply(3:4, function(x)
as.factor(data[[x]]))
# date variables
data <- data %>%
mutate(START_DATE = lubridate::mdy(data$START_DATE),
END_DATE = lubridate::mdy(data$END_DATE))
head(data)## ISO3 COUNTRY_NAME POLICY_TYPE POLICY_SUBTYPE START_DATE END_DATE Country.x
## 1 MAR Morocco COMPLETE ESSENTIAL_ONLY 2020-03-20 2020-06-01 Morocco
## 2 JOR Jordan COMPLETE ESSENTIAL_ONLY 2020-03-17 2020-07-01 Jordan
## 3 LBN Lebanon COMPLETE ESSENTIAL_ONLY 2020-03-18 2020-03-29 Lebanon
## 4 JOR Jordan COMPLETE ESSENTIAL_ONLY 2020-03-17 2020-07-04 Jordan
## 5 MYS Malaysia COMPLETE ESSENTIAL_ONLY 2020-03-18 2020-12-31 Malaysia
## 6 ECU Ecuador COMPLETE ESSENTIAL_ONLY 2020-03-16 2020-04-03 Ecuador
## cases/1000 tests/100 critical case rate mortalityrate Score beds
## 1 0.23341156 1.054939 0.0008130081 1.2419196 -0.11922524 1.00
## 2 0.08972411 2.521581 0.0054644809 0.6849167 0.09947027 1.47
## 3 0.20829699 1.514474 0.0070323488 1.1568900 -0.83280247 2.73
## 4 0.08972411 2.521581 0.0054644809 0.6849167 0.09947027 1.47
## 5 0.25976951 1.911819 0.0004760771 0.8822496 0.99868566 1.88
## 6 2.59681161 0.727088 0.0047839574 2.2366691 -0.39858419 1.39
## population_65_percent death_communicable_disease logisics_score
## 1 7.300455 14.0 2.43
## 2 3.893416 10.7 2.72
## 3 7.273292 3.6 2.64
## 4 3.893416 10.7 2.72
## 5 6.920771 17.5 3.15
## 6 7.372199 15.1 2.72
“In the multiple regression analysis, Covid-19 mortality rate was regressed on Covid-19 test number, case number, critical case rate, government effectiveness score, proportion of population aged 65 or older, number of beds, deaths attributable to communicable diseases, and transport infrastructure quality score.” (Liang, et al, p.2)
# multiple regression
model1 <- lm(mortalityrate ~ `tests/100` + `cases/1000` + `critical case rate` +
`Score` + `population_65_percent` + beds + `death_communicable_disease`+
logisics_score,
data = data)
summary(model1)##
## Call:
## lm(formula = mortalityrate ~ `tests/100` + `cases/1000` + `critical case rate` +
## Score + population_65_percent + beds + death_communicable_disease +
## logisics_score, data = data)
##
## Residuals:
## Min 1Q Median 3Q Max
## -0.91807 -0.30193 0.00226 0.19645 1.62130
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) -0.561507 0.163084 -3.443 0.000618 ***
## `tests/100` -0.028236 0.005801 -4.868 1.46e-06 ***
## `cases/1000` 0.007634 0.007639 0.999 0.318023
## `critical case rate` 19.216809 3.414672 5.628 2.87e-08 ***
## Score -0.383603 0.043651 -8.788 < 2e-16 ***
## population_65_percent 0.098013 0.005051 19.406 < 2e-16 ***
## beds -0.088522 0.009393 -9.424 < 2e-16 ***
## death_communicable_disease 0.000622 0.003221 0.193 0.846952
## logisics_score 0.385968 0.056362 6.848 1.95e-11 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 0.4206 on 568 degrees of freedom
## (13 observations deleted due to missingness)
## Multiple R-squared: 0.5791, Adjusted R-squared: 0.5732
## F-statistic: 97.68 on 8 and 568 DF, p-value: < 2.2e-16
A multiple linear regression was run to examine the relationship between mortality rate and border closure policy type after controlling for Covid-19 test number, case number, critical case rate, government effectiveness score, proportion of the population aged 65 or older, number of beds, deaths attributable to communicable diseases, and transport infrastructure quality score. Changing POLICY_TYPE from COMPLETE to PARTIAl significantly lower the mortality rate by 0.087 (sd = 0.04), t(567) = -2.1, p = 0.03.
# POLICY_TYPE
model_policytype <- lm(mortalityrate ~ POLICY_TYPE + `tests/100` + `cases/1000` + `critical case rate` +
`Score` + `population_65_percent` + beds + `death_communicable_disease`+
logisics_score,
data = data)
summary(model_policytype)##
## Call:
## lm(formula = mortalityrate ~ POLICY_TYPE + `tests/100` + `cases/1000` +
## `critical case rate` + Score + population_65_percent + beds +
## death_communicable_disease + logisics_score, data = data)
##
## Residuals:
## Min 1Q Median 3Q Max
## -0.9902 -0.2912 -0.0076 0.2115 1.6489
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) -0.538338 0.162949 -3.304 0.00101 **
## POLICY_TYPEPARTIAL -0.086721 0.040839 -2.123 0.03415 *
## `tests/100` -0.026683 0.005829 -4.578 5.78e-06 ***
## `cases/1000` 0.006861 0.007624 0.900 0.36856
## `critical case rate` 19.288450 3.404340 5.666 2.33e-08 ***
## Score -0.391017 0.043657 -8.957 < 2e-16 ***
## population_65_percent 0.097187 0.005050 19.245 < 2e-16 ***
## beds -0.086892 0.009396 -9.248 < 2e-16 ***
## death_communicable_disease 0.001104 0.003219 0.343 0.73170
## logisics_score 0.398998 0.056523 7.059 4.92e-12 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 0.4193 on 567 degrees of freedom
## (13 observations deleted due to missingness)
## Multiple R-squared: 0.5824, Adjusted R-squared: 0.5758
## F-statistic: 87.87 on 9 and 567 DF, p-value: < 2.2e-16
When considering only weeks in which more than 50 policies were issued per week worldwide, the above effect of POLICY_TYPE changed its direction, but was no longer significant. b1 = 0.01, sd = 0.07, t(198) = 0.16, p = 0.87.
# during the weeks that had more than 50 policies issued/ week -> 214 policies
week_morethan50 <- data %>%
add_count(week = lubridate::floor_date(START_DATE, "week")) %>%
arrange(desc(n)) %>%
filter(n > 50)
model_policytype_week_morethan50 <- lm(mortalityrate ~ POLICY_TYPE + `tests/100` + `cases/1000` + `critical case rate` + `Score` + `population_65_percent` + beds + `death_communicable_disease`+
logisics_score,
data = week_morethan50)
summary(model_policytype_week_morethan50)##
## Call:
## lm(formula = mortalityrate ~ POLICY_TYPE + `tests/100` + `cases/1000` +
## `critical case rate` + Score + population_65_percent + beds +
## death_communicable_disease + logisics_score, data = week_morethan50)
##
## Residuals:
## Min 1Q Median 3Q Max
## -1.09681 -0.26637 -0.01693 0.21981 1.58711
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) -0.444572 0.308999 -1.439 0.151801
## POLICY_TYPEPARTIAL 0.011751 0.072562 0.162 0.871514
## `tests/100` -0.029184 0.010176 -2.868 0.004578 **
## `cases/1000` 0.002656 0.012571 0.211 0.832880
## `critical case rate` 10.387812 4.726648 2.198 0.029129 *
## Score -0.303241 0.085281 -3.556 0.000471 ***
## population_65_percent 0.099619 0.008542 11.662 < 2e-16 ***
## beds -0.095097 0.017932 -5.303 3.03e-07 ***
## death_communicable_disease 0.004909 0.005479 0.896 0.371351
## logisics_score 0.339910 0.109482 3.105 0.002184 **
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 0.439 on 198 degrees of freedom
## (6 observations deleted due to missingness)
## Multiple R-squared: 0.5739, Adjusted R-squared: 0.5545
## F-statistic: 29.63 on 9 and 198 DF, p-value: < 2.2e-16