Defeat BMI and Psychological Variables
2025-01-15
first we start with loading the data
Then we explore the variables of weight and height
hist(my_data$DM21_01,
main = "Histogram of DM21_01",
xlab = "Weight before",
ylab = "Frequency",
col = "lightblue",
border = "black")
hist(my_data$DM21_02,
main = "Histogram of DM21_02",
xlab = "Weight after",
ylab = "Frequency",
col = "lightblue",
border = "black")
hist(my_data$DM21_03,
main = "Histogram of DM21_03",
xlab = "Height",
ylab = "Frequency",
col = "lightblue",
border = "black")
then We clean data by removing participant with impossible weight and height
my_data_clean <- my_data[my_data$DM21_03 > 60 | is.na(my_data$DM21_03), ]
hist(my_data_clean$DM21_03,
main = "Histogram of DM21_03",
xlab = "Height",
ylab = "Frequency",
col = "lightblue",
border = "black")
then we calculate BMI before and BMI after for each participant as well as BMI change
# Calculate BMI before (based on DM21_01 - weight before and DM21_03 - height)
my_data_clean$BMI_before <- my_data_clean$DM21_01 / (my_data_clean$DM21_03 / 100)^2
# Calculate BMI after (based on DM21_02 - weight after and DM21_03 - height)
my_data_clean$BMI_after <- my_data_clean$DM21_02 / (my_data_clean$DM21_03 / 100)^2
# Display the first few rows to verify the calculations
head(my_data_clean[, c("DM21_01", "DM21_02", "DM21_03", "BMI_before", "BMI_after")])## # A tibble: 6 × 5
## DM21_01 DM21_02 DM21_03 BMI_before BMI_after
## <dbl> <dbl> <dbl> <dbl> <dbl>
## 1 NA NA NA NA NA
## 2 75 75 179 23.4 23.4
## 3 70 65 168 24.8 23.0
## 4 112 125 177 35.7 39.9
## 5 120 120 165 44.1 44.1
## 6 NA NA NA NA NA# Calculate BMI change
my_data_clean$BMI_change <- my_data_clean$BMI_after - my_data_clean$BMI_before
my_data_clean$BMI_change <- as.numeric(as.character(my_data_clean$BMI_change))
# Histogram of BMI changes
hist(my_data_clean$BMI_change,
main = "Histogram of BMI Changes",
xlab = "BMI Change (After - Before)",
col = "lightblue",
border = "black")
Then we check the EQ5D-VAS variable (Gesundheitszustand: Ihr heutiger Gesundheitszustand) and clean data from missing values
my_data_clean$EQ_08_01 <- as.numeric(as.character(my_data_clean$EQ_08_01))
table(my_data_clean$EQ_08_01)##
## -9 1 5 6 7 8 9 10 11 12 13 14 16 17 18 19 20 21 22 23
## 5 1 1 1 2 1 2 1 5 1 1 2 5 7 9 2 3 8 5 4
## 24 25 26 27 28 29 31 32 33 34 35 36 37 38 39 40 41 42 43 44
## 2 6 10 3 2 1 8 7 9 18 11 9 3 3 1 3 14 7 9 5
## 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 65
## 3 5 1 3 2 6 31 3 2 1 1 3 3 2 4 3 8 5 1 2
## 66 67 68 69 70 71 72 73 74 75 76 77 79 80 81 82 83 84 85 86
## 5 13 4 5 5 5 1 2 3 2 5 1 1 2 4 2 5 6 6 3
## 87 89 90 91 92 96 97 100 101
## 4 1 3 1 2 5 2 2 6my_data_clean <- my_data_clean[my_data_clean$EQ_08_01 > 0, ]then first we look to a scatter plot in order to visualize the relationship between BMI change and EQ5D-VAS variable (heutiger Gesundheitszustand) (also I added a linear regression line to the plot so we can take an impression of the relationship between variables)
EQ5D-VAS current score
BMI before vs EQ5D-VAS current
now we are looking for correlation of BMI score of before Covid and current EQ5D-VAS score
plot(my_data_clean$EQ_08_01, my_data_clean$BMI_before,
main = "Scatter Plot of Gesundheitszustand current vs. BMI before)",
xlab = "Gesundheitszustand current",
ylab = "BMI before",
col = "blue", pch = 16)
abline(lm(BMI_before ~ EQ_08_01, data = my_data_clean), col = "red", lwd = 2)
cor_test_BMIbefore_EQ5D_after <- cor.test(my_data_clean$EQ_08_01, my_data_clean$BMI_before, method = "pearson", use = "complete.obs")
print(cor_test_BMIbefore_EQ5D_after)##
## Pearson's product-moment correlation
##
## data: my_data_clean$EQ_08_01 and my_data_clean$BMI_before
## t = -1.2134, df = 381, p-value = 0.2257
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
## -0.1612464 0.0384031
## sample estimates:
## cor
## -0.06204225as you can see there no statistically significant correlation between BMI score before covid and current EQ5D-VAS score
BMI after vs EQ5D-VAS current
plot(my_data_clean$EQ_08_01, my_data_clean$BMI_after,
main = "Scatter Plot of Gesundheitszustand current vs. BMI after)",
xlab = "Gesundheitszustand current",
ylab = "BMI after",
col = "blue", pch = 16)
abline(lm(BMI_after ~ EQ_08_01, data = my_data_clean), col = "red", lwd = 2)
cor_test_BMIafter_EQ5D_after <- cor.test(my_data_clean$EQ_08_01, my_data_clean$BMI_after, method = "pearson", use = "complete.obs")
print(cor_test_BMIafter_EQ5D_after)##
## Pearson's product-moment correlation
##
## data: my_data_clean$EQ_08_01 and my_data_clean$BMI_after
## t = -1.2441, df = 380, p-value = 0.2142
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
## -0.16298800 0.03688173
## sample estimates:
## cor
## -0.06369179BMI change vs EQ5D-VAS Current
plot(my_data_clean$EQ_08_01, my_data_clean$BMI_change,
main = "Scatter Plot of Gesundheitszustand vs. BMI Change",
xlab = "Gesundheitszustand",
ylab = "BMI Change",
col = "blue", pch = 16)
abline(lm(BMI_change ~ EQ_08_01, data = my_data_clean), col = "red", lwd = 2)
Then we continue with calculating pearson correlation of BMI change and and EQ5D-VAS (heutiger Gesundheitszustand)
cor_test <- cor.test(my_data_clean$EQ_08_01, my_data_clean$BMI_change, method = "pearson", use = "complete.obs")
print(cor_test)##
## Pearson's product-moment correlation
##
## data: my_data_clean$EQ_08_01 and my_data_clean$BMI_change
## t = -0.39603, df = 380, p-value = 0.6923
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
## -0.12040409 0.08018958
## sample estimates:
## cor
## -0.02031166So from this test we can conclude there is no direct correlation between BMI change and EQ5D-VAS (Gesundheitszustand)
EQ5D-VAS score change (before corona and current)
then we are looking for the correlation between changes in BMI and changes in EQ5D-VAS status between pre-Covid and now
my_data_clean <- my_data_clean[my_data_clean$DM20_01 > 0, ]
my_data_clean$DM20_01 <- as.numeric(as.character(my_data_clean$DM20_01))
my_data_clean$health_status_change <- my_data_clean$EQ_08_01 - my_data_clean$DM20_01
hist(my_data_clean$health_status_change,
main = "Histogram of health status Changes",
xlab = "health status (Now - before Covid)",
col = "lightblue",
border = "black")
As we can guess participants reported their health condition went worse after Covid,
BMI before vs EQ5D-VAS change
Lets look what will happen if we look on correlation of BMI before Covid and changes in EQ5D-VAS status between pre-Covid and now
plot(my_data_clean$health_status_change, my_data_clean$BMI_before,
main = "Scatter Plot of Gesundheitszustand change vs. BMI before)",
xlab = "Gesundheitszustand change",
ylab = "BMI before",
col = "blue", pch = 16)
abline(lm(BMI_before ~ health_status_change, data = my_data_clean), col = "red", lwd = 2)
cor_test_before <- cor.test(my_data_clean$health_status_change, my_data_clean$BMI_before, method = "pearson", use = "complete.obs")
print(cor_test_before)##
## Pearson's product-moment correlation
##
## data: my_data_clean$health_status_change and my_data_clean$BMI_before
## t = 1.4715, df = 359, p-value = 0.142
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
## -0.02599819 0.17921409
## sample estimates:
## cor
## 0.07742798here we couldn’t spot any fatalistically significant correlation between BMI before covid and change in EQ5D-VAS score changes
BMI after vs EQ5D-VAS change
plot(my_data_clean$health_status_change, my_data_clean$BMI_after,
main = "Scatter Plot of Gesundheitszustand change vs. BMI after)",
xlab = "Gesundheitszustand change",
ylab = "BMI after",
col = "blue", pch = 16)
abline(lm(BMI_after ~ health_status_change, data = my_data_clean), col = "red", lwd = 2)
cor_test_after <- cor.test(my_data_clean$health_status_change, my_data_clean$BMI_after, method = "pearson", use = "complete.obs")
print(cor_test_after)##
## Pearson's product-moment correlation
##
## data: my_data_clean$health_status_change and my_data_clean$BMI_after
## t = 1.1642, df = 358, p-value = 0.2451
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
## -0.04221429 0.16373792
## sample estimates:
## cor
## 0.06141551lets look if there is a correlation between change of their health condition and their BMI during the period
BMI change vs EQ5D-VAS change
plot(my_data_clean$health_status_change, my_data_clean$BMI_change,
main = "Scatter Plot of Gesundheitszustand change vs. BMI Change)",
xlab = "Gesundheitszustand change",
ylab = "BMI Change",
col = "blue", pch = 16)
abline(lm(BMI_change ~ health_status_change, data = my_data_clean), col = "red", lwd = 2)
cor_test_HC <- cor.test(my_data_clean$health_status_change, my_data_clean$BMI_change, method = "pearson", use = "complete.obs")
print(cor_test_HC)##
## Pearson's product-moment correlation
##
## data: my_data_clean$health_status_change and my_data_clean$BMI_change
## t = -0.62203, df = 358, p-value = 0.5343
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
## -0.1357585 0.0707443
## sample estimates:
## cor
## -0.03285779Again no correlation found between the two variables.
lets dive into IMET scale, I am first starting with calculating total IMET score based on question of IMET 1 to 8 (I only include participants who have answerd all of the questions)
# Variables of interest
variables_to_convert <- paste0("ME01_0", 1:8)
# Convert to numeric, replace -9 with NA, and recode values from 1–11 to 0–10
my_data_clean[variables_to_convert] <- my_data_clean[variables_to_convert] |>
lapply(function(column) {
numeric_column <- as.numeric(column) # Convert to numeric
numeric_column[numeric_column == -9] <- NA # Replace -9 with NA
numeric_column <- numeric_column - 1 # Recode values: 1–11 to 0–10
return(numeric_column)
}) |>
as.data.frame() # Convert back to data frame
# Create a new variable by summing the values of the selected columns,
# setting NA for participants with any missing values in the selected variables
my_data_clean$ME01_sum <- apply(my_data_clean[variables_to_convert], 1, function(row) {
if (any(is.na(row))) {
return(NA) # Set NA if any variable is missing
} else {
return(sum(row)) # Calculate the sum if all variables have values
}
})IMET standard (only 8 main questions)
BMI before vs IMET standard
plot(my_data_clean$ME01_sum, my_data_clean$BMI_before,
main = "Scatter Plot of Gesundheitszustand vs. BMI before",
xlab = "IMET total",
ylab = "BMI before",
col = "blue", pch = 16)
abline(lm(BMI_before ~ ME01_sum, data = my_data_clean), col = "red", lwd = 2)
cor_test_IMET <- cor.test(my_data_clean$ME01_sum, my_data_clean$BMI_before, method = "pearson", use = "complete.obs")
print(cor_test_IMET)##
## Pearson's product-moment correlation
##
## data: my_data_clean$ME01_sum and my_data_clean$BMI_before
## t = 2.7312, df = 345, p-value = 0.006634
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
## 0.04082206 0.24697911
## sample estimates:
## cor
## 0.1454794The Pearson correlation analysis shows a weak positive relationship between ME01_sum and BMI_before (r = 0.145, p = 0.007). The p-value indicates that the correlation is statistically significant at the 0.05 level, suggesting that as BMI_before increases, there is a slight tendency for ME01_sum to increase.(participants who had a higher BMI before covid now have a worse IMET score)
BMI after vs IMET standard
plot(my_data_clean$ME01_sum, my_data_clean$BMI_after,
main = "Scatter Plot of Gesundheitszustand vs. BMI after",
xlab = "IMET total",
ylab = "BMI after",
col = "blue", pch = 16)
abline(lm(BMI_after ~ ME01_sum, data = my_data_clean), col = "red", lwd = 2)
cor_test_IMET2 <- cor.test(my_data_clean$ME01_sum, my_data_clean$BMI_after, method = "pearson", use = "complete.obs")
print(cor_test_IMET2)##
## Pearson's product-moment correlation
##
## data: my_data_clean$ME01_sum and my_data_clean$BMI_after
## t = 2.7936, df = 344, p-value = 0.005504
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
## 0.04420032 0.25044296
## sample estimates:
## cor
## 0.148941again we are seeing a similar pattern here for BMI after total IMET score
BMI change vs IMET standard
plot(my_data_clean$ME01_sum, my_data_clean$BMI_change,
main = "Scatter Plot of Gesundheitszustand vs. BMI Change",
xlab = "IMET total",
ylab = "BMI Change",
col = "blue", pch = 16)
abline(lm(BMI_change ~ ME01_sum, data = my_data_clean), col = "red", lwd = 2)
cor_test_IMET3 <- cor.test(my_data_clean$ME01_sum, my_data_clean$BMI_change, method = "pearson", use = "complete.obs")
print(cor_test_IMET3)##
## Pearson's product-moment correlation
##
## data: my_data_clean$ME01_sum and my_data_clean$BMI_change
## t = 0.64114, df = 344, p-value = 0.5219
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
## -0.07114663 0.13947408
## sample estimates:
## cor
## 0.03454732But for BMI change we don’t have any statistical significant correlation between total IMET score and BMI change
# Load dplyr for recode function
# Step 1: Dynamically find columns matching "FA01_01" to "FA01_10"
variables_to_convert2 <- grep("^FA01_0[1-9]$|^FA01_10$", names(my_data_clean), value = TRUE)
# Step 2: Convert to numeric and replace -9 with NA
my_data_clean[variables_to_convert2] <- as.data.frame(
lapply(my_data_clean[variables_to_convert2], function(column) {
numeric_column <- as.numeric(as.character(column)) # Convert to numeric
numeric_column[numeric_column == -9] <- NA # Replace -9 with NA
return(numeric_column)
})
)
# Step 3: Verify the changes
# Check the selected variables
# Confirm numeric conversion
# Step 3: Recode scores for questions 4 and 10
my_data_clean$FA01_04 <- recode(my_data_clean$FA01_04, `1` = 5, `2` = 4, `3` = 3, `4` = 2, `5` = 1)
my_data_clean$FA01_10 <- recode(my_data_clean$FA01_10, `1` = 5, `2` = 4, `3` = 3, `4` = 2, `5` = 1)
# Step 4: Calculate the total FAS score only for participants who answered all questions
my_data_clean$FAS_total <- apply(my_data_clean[variables_to_convert2], 1, function(row) {
if (any(is.na(row))) {
return(NA) # Set NA if any question is unanswered
} else {
return(sum(row)) # Calculate the sum if all questions are answered
}
})FAS
BMI before vs FAS
plot(my_data_clean$FAS_total, my_data_clean$BMI_before,
main = "Scatter Plot of FAS total vs. BMI before",
xlab = "FAS_total",
ylab = "BMI before",
col = "blue", pch = 16)
abline(lm(BMI_before ~ FAS_total, data = my_data_clean), col = "red", lwd = 2)
cor_test_FAS_BMIbefore <- cor.test(my_data_clean$FAS_total, my_data_clean$BMI_before, method = "pearson", use = "complete.obs")
print(cor_test_FAS_BMIbefore)##
## Pearson's product-moment correlation
##
## data: my_data_clean$FAS_total and my_data_clean$BMI_before
## t = 3.5779, df = 340, p-value = 0.0003966
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
## 0.08617858 0.29066652
## sample estimates:
## cor
## 0.190488we found a weak positive relationship between BMI before corona and FAS score (r = 0.190, p = 0.0004). The significant p-value and 95% CI (0.086, 0.291) confirm a small but positive association.
BMI after vs FAS
plot(my_data_clean$FAS_total, my_data_clean$BMI_after,
main = "Scatter Plot of FAS total vs. BMI after",
xlab = "FAS_total",
ylab = "BMI after",
col = "blue", pch = 16)
abline(lm(BMI_after ~ FAS_total, data = my_data_clean), col = "red", lwd = 2)
cor_test_FAS_BMIafter <- cor.test(my_data_clean$FAS_total, my_data_clean$BMI_after, method = "pearson", use = "complete.obs")
print(cor_test_FAS_BMIafter)##
## Pearson's product-moment correlation
##
## data: my_data_clean$FAS_total and my_data_clean$BMI_after
## t = 3.7274, df = 339, p-value = 0.0002267
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
## 0.09419661 0.29833659
## sample estimates:
## cor
## 0.1984175Also there is a weak but statistically significant correlation between BMI after corona and FAS score
BMI change vs FAS
plot(my_data_clean$FAS_total, my_data_clean$BMI_change,
main = "Scatter Plot of FAS total vs. BMI change",
xlab = "FAS_total",
ylab = "BMI change",
col = "blue", pch = 16)
abline(lm(BMI_change ~ FAS_total, data = my_data_clean), col = "red", lwd = 2)
cor_test_FAS_BMIchange <- cor.test(my_data_clean$FAS_total, my_data_clean$BMI_change, method = "pearson", use = "complete.obs")
print(cor_test_FAS_BMIchange)##
## Pearson's product-moment correlation
##
## data: my_data_clean$FAS_total and my_data_clean$BMI_change
## t = 1.1044, df = 339, p-value = 0.2702
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
## -0.04662708 0.16503184
## sample estimates:
## cor
## 0.05987536there is no he statistically significant correlation between BMI change and FAS score
IMET vs EQ5D-VAS current
at the end I wanted to look up if there is a strong correlation between IMET total variable and EQ5D-VAS. as we can expect a strong negative correlation. (higher values for IMET indicates worse health condition and higher values for EQ5D-VAS shows better health conditions)
plot(my_data_clean$EQ_08_01, my_data_clean$ME01_sum,
main = "Scatter Plot of Gesundheitszustand vs. IMET sum",
xlab = "Gesundheitszustand",
ylab = "IMET sum",
col = "blue", pch = 16)
abline(lm(ME01_sum ~ EQ_08_01, data = my_data_clean), col = "red", lwd = 2)
As we can see correlation is as could expect, so we can continue with our research question about relation of BMI (before,after and change) and IMET scale