In this assignment, we analyse the dataset selected from Global Dietary Database website, which includes B12 intake of former Soviet union countries, each row representing the B12 intake estimation for each 5 year between 1990 and 2020 and for each country, based on the variables of gender, residence (urban or rural), age group and education level.
gdd <- read.csv("https://raw.githubusercontent.com/berkorbay/datasets/master/gdd/gdd_b12_levels.csv")
library(dplyr)
Attaching package: 'dplyr'The following objects are masked from 'package:stats':
filter, lagThe following objects are masked from 'package:base':
intersect, setdiff, setequal, unionlibrary(ggplot2)#1.1) Gender
gdd_g <- gdd %>%
mutate(gender = case_when(female == 1 ~ 'female', female == 0 ~ 'male', female == 999 ~ 'all genders'))
gdd_g %>%
head(10) iso3 age female urban edu year median lowerci_95 upperci_95 gender
1 ALB 999 0 999 999 1990 3.634332 2.303403 6.054550 male
2 ALB 999 0 999 999 1995 3.737716 2.433315 5.981505 male
3 ALB 999 0 999 999 2000 3.824925 2.513833 6.216503 male
4 ALB 999 0 999 999 2005 3.938703 2.665851 6.267153 male
5 ALB 999 0 999 999 2010 4.067680 2.738054 6.253339 male
6 ALB 999 0 999 999 2015 4.092445 2.777237 6.344306 male
7 ALB 999 0 999 999 2018 4.070591 2.761252 6.266195 male
8 ALB 999 0 999 999 2020 4.087936 2.801015 6.307502 male
9 ALB 999 1 999 999 1990 3.292622 2.004336 5.514846 female
10 ALB 999 1 999 999 1995 3.399279 2.134920 5.449213 female#1.2) Urban / Rural
gdd_r <- gdd_g %>%
mutate(residence = case_when(urban == 1 ~ 'urban', urban == 0 ~ 'rural', urban == 999 ~ 'all residences'))
gdd_r %>%
head(10) iso3 age female urban edu year median lowerci_95 upperci_95 gender
1 ALB 999 0 999 999 1990 3.634332 2.303403 6.054550 male
2 ALB 999 0 999 999 1995 3.737716 2.433315 5.981505 male
3 ALB 999 0 999 999 2000 3.824925 2.513833 6.216503 male
4 ALB 999 0 999 999 2005 3.938703 2.665851 6.267153 male
5 ALB 999 0 999 999 2010 4.067680 2.738054 6.253339 male
6 ALB 999 0 999 999 2015 4.092445 2.777237 6.344306 male
7 ALB 999 0 999 999 2018 4.070591 2.761252 6.266195 male
8 ALB 999 0 999 999 2020 4.087936 2.801015 6.307502 male
9 ALB 999 1 999 999 1990 3.292622 2.004336 5.514846 female
10 ALB 999 1 999 999 1995 3.399279 2.134920 5.449213 female
residence
1 all residences
2 all residences
3 all residences
4 all residences
5 all residences
6 all residences
7 all residences
8 all residences
9 all residences
10 all residences#1.3) Education Level
gdd_f <- gdd_r %>%
mutate(edu_level = case_when(edu == 1 ~ 'low', edu == 2 ~ 'medium', edu == 3 ~ 'high',edu == 999 ~ 'all levels'))
gdd_f %>%
head(10) iso3 age female urban edu year median lowerci_95 upperci_95 gender
1 ALB 999 0 999 999 1990 3.634332 2.303403 6.054550 male
2 ALB 999 0 999 999 1995 3.737716 2.433315 5.981505 male
3 ALB 999 0 999 999 2000 3.824925 2.513833 6.216503 male
4 ALB 999 0 999 999 2005 3.938703 2.665851 6.267153 male
5 ALB 999 0 999 999 2010 4.067680 2.738054 6.253339 male
6 ALB 999 0 999 999 2015 4.092445 2.777237 6.344306 male
7 ALB 999 0 999 999 2018 4.070591 2.761252 6.266195 male
8 ALB 999 0 999 999 2020 4.087936 2.801015 6.307502 male
9 ALB 999 1 999 999 1990 3.292622 2.004336 5.514846 female
10 ALB 999 1 999 999 1995 3.399279 2.134920 5.449213 female
residence edu_level
1 all residences all levels
2 all residences all levels
3 all residences all levels
4 all residences all levels
5 all residences all levels
6 all residences all levels
7 all residences all levels
8 all residences all levels
9 all residences all levels
10 all residences all levelsIn order to compare the general country B12 intakes in between, filter all the variables as 999 (representing general population) and take the mean and median for each country and summarise:
gdd_1 <- gdd %>%
filter(age== 999, female == 999, urban == 999, edu == 999) %>%
group_by(iso3, year) %>%
summarise(mean_values = (lowerci_95+upperci_95)/2, median) %>%
arrange(desc(median))`summarise()` has grouped output by 'iso3'. You can override using the
`.groups` argument.gdd_1# A tibble: 232 × 4
# Groups: iso3 [29]
iso3 year mean_values median
<chr> <int> <dbl> <dbl>
1 EST 2010 7.16 7.10
2 EST 1990 7.15 7.09
3 EST 2018 7.14 7.08
4 EST 2005 7.13 7.08
5 EST 2020 7.13 7.07
6 EST 2015 7.12 7.07
7 EST 2000 7.10 7.05
8 EST 1995 7.07 7.00
9 BLR 2020 4.50 4.10
10 BLR 2018 4.51 4.07
# ℹ 222 more rowsgdd_2 <- ggplot(gdd_1, aes(x = year, y = median, color = iso3)) + geom_line()
gdd_2
None of the countries show an extreme change in median B12 intake throughout the years, TJK as the most changing one. As per the general population intakes of countries, BGR and partly TJK are the outlier values in terms of median with lower intake. Est has the outlier value as higher intake.It is difficult to talk about a trend per country.
gdd_3 <- ggplot(gdd_1, aes(x = year, y = mean_values, color = iso3)) + geom_line()
gdd_3
For the countries around average mean value, we can see an extreme upward trend from year 2000 to 2010 and then a decline towards 2015, and remain constant between 2015 and 2020.
In terms of mean, there are outliers below 4 mg intake,but hard to recognise as the colors are similar, and consistently with median graph, EST is the upper outlier with higher intake.
To show the general population mean B12 intake with a bar chart for each country, first we summarise the table with the average of all the years rather than for each year:
gdd_4 <- gdd %>%
filter(age== 999, female == 999, urban == 999, edu == 999) %>%
group_by(iso3) %>%
summarise(mean_values = mean(lowerci_95+upperci_95)/2) %>%
arrange(desc(mean_values))
gdd_4# A tibble: 29 × 2
iso3 mean_values
<chr> <dbl>
1 EST 7.12
2 CZE 4.45
3 BLR 4.43
4 RUS 4.39
5 MNG 4.37
6 LTU 4.37
7 LVA 4.36
8 HUN 4.33
9 SVN 4.33
10 MNE 4.32
# ℹ 19 more rows#bar chart to compare country average intake:
gdd_5 <- ggplot(gdd_4, aes(x = iso3, y = mean_values, color = iso3)) + geom_col()
gdd_5
We can see three BGR, ROU and POL have lower average intake and EST as higher intake more clearly in this chart.
We can analyse the countries with outlier values in more detail. To see the effect of age groups for B12 intake, select countries BGR, ROU and POL (lower intake), select for all lifestyles, all education levels, all genders but only different age groups, and visualise the table:
gdd_6 <- gdd %>%
filter(female == 999, urban == 999, edu == 999, age < 999, iso3 == "BGR" | iso3 == "ROU"| iso3 == "POL") %>%
group_by(age) %>%
#again to ensure we take the average of seperate years rather than all the years:
summarise(mean_values = mean(lowerci_95+upperci_95)/2) %>%
arrange(desc(mean_values))
gdd_6# A tibble: 22 × 2
age mean_values
<dbl> <dbl>
1 22.5 3.61
2 27.5 3.60
3 32.5 3.57
4 17.5 3.57
5 37.5 3.53
6 42.5 3.48
7 12.5 3.43
8 47.5 3.42
9 52.5 3.37
10 57.5 3.31
# ℹ 12 more rows#visualise to see if there is any major change by different age group:
gdd_7 <- ggplot(gdd_6, aes(x = age, y = mean_values, fill = age)) + geom_col() + geom_text(aes(label = round(mean_values, 1)))
gdd_7
We can see that B12 intake for children up tp 10 year is very low (ranging from 1.5 mg to 2.5mg), however for age groups taking higher B12, the average intake is roughly above 3. It seems there is no extreme low intake for some age groups affecting the average of these countries.
To see if for lower intake countries the age group intake is similar to the other countries, we can prepare the same chart including all countries and prepare another bar chart:
gdd_8 <- gdd %>%
filter(female == 999, urban == 999, edu == 999, age < 999) %>%
group_by(age) %>%
#again to ensure we take the average of seperate years rather than all the years:
summarise(mean_values = mean(lowerci_95+upperci_95)/2) %>%
arrange(desc(mean_values))
gdd_8# A tibble: 22 × 2
age mean_values
<dbl> <dbl>
1 22.5 4.75
2 27.5 4.74
3 17.5 4.70
4 32.5 4.70
5 37.5 4.64
6 42.5 4.58
7 47.5 4.51
8 52.5 4.44
9 12.5 4.44
10 57.5 4.37
# ℹ 12 more rowsgdd_9 <- ggplot(gdd_8, aes(x = age, y = mean_values, fill=age)) + geom_col() + geom_text(aes(label = round(mean_values, 1)))
gdd_9
It seems the children’s intake trend compared to adolescent and adult intake is in line for lower-average countries when compared to all countries.We can say that the lower intake of a specific group affecting the average is not the case here.
gdd_10 <- gdd_f %>%
filter(age== 999, gender == "female" | gender == "male", residence == "urban" | residence == "rural", edu == 999) %>%
group_by(gender, residence) %>%
#to ensure we take the average of seperate years rather than all the years:
summarise(mean_values = mean(lowerci_95+upperci_95)/2) %>%
arrange(desc(mean_values))`summarise()` has grouped output by 'gender'. You can override using the
`.groups` argument.gdd_10# A tibble: 4 × 3
# Groups: gender [2]
gender residence mean_values
<chr> <chr> <dbl>
1 male urban 4.75
2 female urban 4.27
3 male rural 4.17
4 female rural 3.75gdd_11 <- ggplot(gdd_10, aes(x=residence, y=mean_values, fill=gender)) + geom_bar(stat="identity") + expand_limits(x=0) + expand_limits(y=0) + geom_text(aes(label = round(mean_values, 1)), position = position_stack(vjust = 0.5))
gdd_11
gdd_12 <- gdd_f %>%
filter(age== 999, gender == "female" | gender == "male", residence == "urban" | residence == "rural", edu == 999, iso3 == "BGR" | iso3 == "ROU"| iso3 == "POL") %>%
group_by(gender, residence) %>%
#to ensure we take the average of seperate years rather than all the years:
summarise(mean_values = mean(lowerci_95+upperci_95)/2) %>%
arrange(desc(mean_values))`summarise()` has grouped output by 'gender'. You can override using the
`.groups` argument.gdd_12# A tibble: 4 × 3
# Groups: gender [2]
gender residence mean_values
<chr> <chr> <dbl>
1 male urban 3.84
2 male rural 3.37
3 female urban 3.21
4 female rural 2.82gdd_13 <- ggplot(gdd_12, aes(x=residence, y=mean_values, fill=gender)) + geom_bar(stat="identity") + expand_limits(x=0) + expand_limits(y=0) + geom_text(aes(label = round(mean_values, 1)), position = position_stack(vjust = 0.5))
gdd_13
gdd_14 <- gdd_f %>%
filter(age== 999, gender == "female" | gender == "male", residence == "urban" | residence == "rural", edu == 999, iso3 == "EST") %>%
group_by(gender, residence) %>%
#to ensure we take the average of seperate years rather than all the years:
summarise(mean_values = mean(lowerci_95+upperci_95)/2) %>%
arrange(desc(mean_values))`summarise()` has grouped output by 'gender'. You can override using the
`.groups` argument.gdd_14# A tibble: 4 × 3
# Groups: gender [2]
gender residence mean_values
<chr> <chr> <dbl>
1 male urban 7.69
2 female urban 7.20
3 male rural 6.73
4 female rural 6.30gdd_15 <- ggplot(gdd_14, aes(x=residence, y=mean_values, fill=gender)) + geom_bar(stat="identity") + expand_limits(x=0) + expand_limits(y=0) + geom_text(aes(label = round(mean_values, 1)), position = position_stack(vjust = 0.5))
gdd_15
The gap between the females and males is lowest for EST, high intake country, in relative to the other groups, and relatively the highest for lower intake countries, which means female intake is significantly lower than males in low-intake countries. The gap between rural an urban areas is similar in three groups, urban areas having higher intake.
gdd_16 <- gdd_f %>%
filter(age== 999, urban == 999, edu_level == "low" | edu_level == "medium" | edu_level == "high", gender == "male" | gender == "female") %>%
group_by(edu_level, gender) %>%
#to ensure we take the average of seperate years rather than all the years:
summarise(mean_values = mean(lowerci_95+upperci_95)/2) %>%
arrange(desc(mean_values))`summarise()` has grouped output by 'edu_level'. You can override using the
`.groups` argument.gdd_16# A tibble: 6 × 3
# Groups: edu_level [3]
edu_level gender mean_values
<chr> <chr> <dbl>
1 medium male 4.58
2 high male 4.49
3 low male 4.27
4 medium female 4.12
5 high female 4.04
6 low female 3.84gdd_17 <- ggplot(gdd_16, aes(x=gender, y=mean_values, fill=edu_level)) + geom_bar(stat="identity") + expand_limits(x=0) + expand_limits(y=0) + geom_text(aes(label = round(mean_values, 1)), position = position_stack(vjust = 0.5))
gdd_17
gdd_18 <- gdd_f %>%
filter(age== 999, urban == 999, edu_level == "low" | edu_level == "medium" | edu_level == "high", gender == "male" | gender == "female", iso3 == "BGR" | iso3 == "ROU"| iso3 == "POL") %>%
group_by(edu_level, gender) %>%
#to ensure we take the average of seperate years rather than all the years:
summarise(mean_values = mean(lowerci_95+upperci_95)/2) %>%
arrange(desc(mean_values))`summarise()` has grouped output by 'edu_level'. You can override using the
`.groups` argument.gdd_18# A tibble: 6 × 3
# Groups: edu_level [3]
edu_level gender mean_values
<chr> <chr> <dbl>
1 medium male 3.73
2 high male 3.64
3 low male 3.48
4 medium female 3.12
5 high female 3.05
6 low female 2.90gdd_19 <- ggplot(gdd_18, aes(x=gender, y=mean_values, fill=edu_level)) + geom_bar(stat="identity") + expand_limits(x=0) + expand_limits(y=0) + geom_text(aes(label = round(mean_values, 1)), position = position_stack(vjust = 0.5))
gdd_19
gdd_20 <- gdd_f %>%
filter(age== 999, urban == 999, edu_level == "low" | edu_level == "medium" | edu_level == "high", gender == "male" | gender == "female", iso3 == "EST") %>%
group_by(edu_level, gender) %>%
#to ensure we take the average of seperate years rather than all the years:
summarise(mean_values = mean(lowerci_95+upperci_95)/2) %>%
arrange(desc(mean_values))`summarise()` has grouped output by 'edu_level'. You can override using the
`.groups` argument.gdd_20# A tibble: 6 × 3
# Groups: edu_level [3]
edu_level gender mean_values
<chr> <chr> <dbl>
1 medium male 7.50
2 high male 7.33
3 medium female 7.02
4 low male 6.98
5 high female 6.86
6 low female 6.53gdd_21 <- ggplot(gdd_20, aes(x=gender, y=mean_values, fill=edu_level)) + geom_bar( stat="identity") + expand_limits(x=0) + expand_limits(y=0) + geom_text(aes(label = round(mean_values, 1)), position = position_stack(vjust = 0.5))
gdd_21
No matter the country group, the difference of B12 intake of males vs females does not vary significantly based on the education level.