Is the Media Noise around Re-sale price (2022) right?

Author

Aishwarya Maloo

Published

February 10, 2023

1. The Task

The task is to uncover the salient patterns of the resale prices of public housing property by residential towns and estates in Singapore by using appropriate analytical visualisation techniques learned in Lesson 4: Fundamentals of Visual Analytics. Students are encouraged to apply appropriate interactive techniques to enhance user and data discovery experiences.

For the purpose of this study, the focus should be on 3-ROOM, 4-ROOM and 5-ROOM types. You can choose to focus on either one housing type or multiple housing types. The study period should be on 2022.

2. Description of Dataset

The data used is ’Resale flat princes based on registration date from Jan-2017 onwards’, which was obtained from Data.gov.sg. The data has 147601 rows and 11 columns.

Columns Description
month Month and year of listing
town Planning area in Singapore
flat_type Type of flats available - 1 room, 2 room, 3 room, etc
block Block number of flat
street_name Name of the street where the flat is located
storey_range Range of storey
floor_area_sqm The area of the flat in square metre
flat_model Model of flat type - new generation, model A, etc
lease_commence_date Year the lease started
remaining_lease Number of years and months pending on lease lapse
resale_price Total resale price of flat

3. Data Wrangling and Preparation

3.1 Install Requisite R Packages

pacman::p_load(ggstatsplot, ggthemes, plotly, corrplot, lubridate, ggpubr, plotly, treemap, d3treeR, hrbrthemes, ggrepel, RColorBrewer, gganimate, viridis, ggridges, ggrepel, testthat, hmisc, tidyverse)

3.2 Loading the Resale Flat - Singapore Dataset

sg_re_price <- read_csv("data/resale_flat_prices.csv")

3.3 Data Prep

Issues Description Resolution
Outliers The resale price has outliers where the price goes up 141800. The top 10% of resale price are outliers. Outliers should be excluded from the data. Anything falling above the 3rd quartile is filtered out from the data. (3.3.4)
Variable type Date is available in the form of ‘YYYY - MM’ which is not readable in R Format date, seperate month into month and year. (3.3.1)
New variables Lease start date and lease years cannot be used for analysis. Additionally, floor area and reslae price can be used to calculate price per square meter to enhance analysis. price_psm, priceK and property age are calculated. (3.3.2)

3.3.1. Separating the date into month and year

Code 
sg_re_price <- sg_re_price %>% 
  separate(month, c("Year", "Month"), sep = "-")
Code 
sg_re_price$Year <- strtoi(sg_re_price$Year)
sg_re_price$Month <- strtoi(sg_re_price$Month)

3.3.2 Creating new variables

Code 
sg_re_price <- sg_re_price %>%
  mutate(price_psm = round(resale_price / floor_area_sqm)) %>%
  mutate(priceK = round(resale_price / 1000)) %>%
  mutate(property_age = round(2022 - lease_commence_date))

3.3.3 Considering data for year 2022 and 3-room, 4-room and 5-room flat types

Code 
sg_re_price_2022 <- sg_re_price %>%
  filter(Year == 2022, flat_type %in% c("3 ROOM", "4 ROOM", "5 ROOM"))

3.3.4 Summary of Data

The filtered data has 24, 372 observations. Please find below the summary statistics for each of the variables.

psych::describe(sg_re_price_2022)
                    vars     n      mean        sd median   trimmed       mad
Year                   1 24372   2022.00      0.00   2022   2022.00      0.00
Month                  2 19922      6.05      3.65      5      5.93      2.97
town*                  3 24372     15.12      7.88     17     15.47      8.90
flat_type*             4 24372      2.02      0.73      2      2.02      1.48
block*                 5 24372   1059.67    706.08   1018   1035.46    886.59
street_name*           6 24372    275.03    165.99    275    274.60    212.01
storey_range*          7 24372      3.36      2.09      3      3.08      1.48
floor_area_sqm         8 24372     94.08     19.32     93     93.99     25.20
flat_model*            9 24372      6.78      2.63      6      6.52      2.97
lease_commence_date   10 24372   1997.46     14.98   1998   1997.83     20.76
remaining_lease*      11 24372    376.27    180.22    377    380.97    247.59
resale_price          12 24372 536394.11 157999.71 515000 520463.70 133434.00
price_psm             13 24372   5735.91   1363.09   5368   5511.74    913.28
priceK                14 24372    536.40    158.00    515    520.47    133.43
property_age          15 24372     24.54     14.98     24     24.17     20.76
                       min     max   range  skew kurtosis      se
Year                  2022    2022       0   NaN      NaN    0.00
Month                    1      12      11  0.31    -1.21    0.03
town*                    1      26      25 -0.31    -1.16    0.05
flat_type*               1       3       2 -0.02    -1.13    0.00
block*                   1    2457    2456  0.21    -1.16    4.52
street_name*             1     552     551  0.02    -1.26    1.06
storey_range*            1      17      16  1.66     4.36    0.01
floor_area_sqm          51     159     108 -0.06    -0.85    0.12
flat_model*              1      16      15  0.71    -0.39    0.02
lease_commence_date   1967    2019      52 -0.03    -1.33    0.10
remaining_lease*         1     638     637 -0.04    -1.32    1.15
resale_price        200000 1418000 1218000  1.09     1.81 1012.07
price_psm             3333   14731   11398  1.70     3.30    8.73
priceK                 200    1418    1218  1.09     1.81    1.01
property_age             3      55      52  0.03    -1.33    0.10

3.3.5. Historgram and dealing with outliers

To check the normality of data, we chart out histograms for resale price (in 1000s), Floor area, Property age and price per square meter. We observe that there is a distinct right skew for price per square and resale price ((in 1000s). Our aim is for the data to be normal or as normal as possible.

Code 
set.seed(1234)

#need to change bar colors, line color, ggtitles, gglabs

p1 <- gghistostats(
  data = sg_re_price_2022,
  x = priceK,
  type = "bayes",
  test.value = 60,
  xlab = "Resale Price in Thousands") +
  
  theme_minimal() +
  
  theme(text = element_text(family = "Garamond"))
        
p2 <- gghistostats(
  data = sg_re_price_2022,
  x = floor_area_sqm,
  type = "bayes",
  test.value = 60,
  xlab = "Floor area (sqm)"
) +
  theme_minimal() +
  
  theme(text = element_text(family = "Garamond"))

p3 <- gghistostats(
  data = sg_re_price_2022,
  x = property_age,
  type = "bayes",
  test.value = 60,
  xlab = "Property Age"
) +
  theme_minimal() +
  
  theme(text = element_text(family = "Garamond"))

p4 <- gghistostats(
  data = sg_re_price_2022,
  x = price_psm,
  type = "bayes",
  test.value = 60,
  xlab = "Resale price psm"
) +
  theme_minimal()+
  
  theme(text = element_text(family = "Garamond"))

3. Remove outliers: The rule of thumb for outliers is that data points over the third quartile are eliminated. We calculate the upper limit and interquartile range for both the variables and then filter out the outliers.

Code 
# Calculating the upper limit and Interquartile range for Price in 1000s and price per square meter

IQR_priceK = IQR(sg_re_price_2022$priceK)
IQR_price_psm = IQR(sg_re_price_2022$price_psm)

priceK_upper = quantile(sg_re_price_2022$priceK,probs = 0.9)+1.5*IQR_priceK
price_psm_upper = quantile(sg_re_price_2022$price_psm,probs = 0.9)+1.5*IQR_price_psm

# Filtering out the outliers
sg_re_price_2022_v1 <- sg_re_price_2022 %>%
           filter ((priceK <= priceK_upper) &
           (price_psm<= price_psm_upper))

4. View data after outliers are removed

Code 
#To plot normality, we need to ascertain the mean and std. deviation

mean_priceK = mean(sg_re_price_2022_v1$priceK)
std_priceK = sd(sg_re_price_2022_v1$priceK)

priceK_norm <- ggplot(sg_re_price_2022_v1, aes(priceK))+
  geom_histogram(aes(y=..density..), fill = '#133337', color = '#eeeeee')+
  
  stat_function(fun = dnorm, args = list(mean = mean_priceK, sd = std_priceK), col="#66cccc", size = .7)+
  
  labs(title = 'Normality Check on Distribution of Resale Price', 
       x = "Resale price in Thousands (SGD)")+
  
  theme_minimal()+
  
  theme(text = element_text(family = "Garamond"),
        plot.title = element_text(hjust = 0.1, size = 10, face = 'bold'),
        plot.margin = margin(20, 20, 20, 20),
        
        panel.background = element_rect(fill = 'white'),
        panel.grid.major.x = element_line(linewidth = 0.25, linetype = 'dashed', colour = '#bebebe'),
        panel.grid.major.y = element_blank(),
        panel.grid.minor.x = element_blank(),
        panel.grid.minor.y = element_blank(),
        
        
        axis.text.y = element_blank(),
        axis.title.y = element_blank(),
        axis.ticks.y = element_blank(),
        axis.text.x = element_text(size = 8, face = "bold"),
        axis.title.x = element_text(hjust = 0.4, size = 10, face = "bold"))
Code 
#To plot normality, we need to ascertain the mean and std. deviation

mean_price_psm = mean(sg_re_price_2022_v1$price_psm)
std_price_psm = sd(sg_re_price_2022_v1$price_psm)

price_per_psm_norm <- ggplot(sg_re_price_2022_v1, aes(price_psm))+
  geom_histogram(aes(y=..density..), fill = '#133337', color = '#eeeeee')+
  stat_function(fun = dnorm, args = list(mean = mean_price_psm, sd = std_price_psm), col="#66cccc", size = .7)+
  
  labs(title = 'Normality Check on Distribution of Resale Price psm', 
       x = "Resale price per sqm (SGD)", 
       caption = "Source: Data.gov.sg")+
  
  theme_minimal()+
  
  theme(text = element_text(family = "Garamond"),
        plot.title = element_text(hjust = 0.1, size = 10, face = 'bold'),
        plot.margin = margin(20, 20, 20, 20),
        plot.caption = element_text(face = "italic", size = 10),
        
        panel.background = element_rect(fill = 'white'),
        panel.grid.major.x = element_line(linewidth = 0.25, linetype = 'dashed', colour = '#bebebe'),
        panel.grid.major.y = element_blank(),
        panel.grid.minor.x = element_blank(),
        panel.grid.minor.y = element_blank(),
        
        
        axis.text.y = element_blank(),
        axis.title.y = element_blank(),
        axis.ticks.y = element_blank(),
        axis.text.x = element_text(size = 8, face = "bold"),
        axis.title.x = element_text(hjust = 0.5, size = 10, face = "bold"))
Code 
priceK_norm + price_per_psm_norm

The distribution for resale price and resale price psm appears to be less right skewed post removing outliers!

4. Data Visualisation

4.1 Flat Type

4.1.1 Resale prices over the years by flat type

Ridgeline plot is a set of overlapped density plots, and it helps us to compare multiple distirbutions among dataset. They are useful for visualizing changes in distributions over time or space.

From this graph we learn:

  1. The resale prices are more or less contained from 2017 to 2020. There is a marked increase (sharp right movement of all curves) in prices for all flat types in 2020, then a gradual increase is observed from 2021.
  2. There is a slow increase in prices observed over 2022. The distribution of 5-room flats observes rapid fluctuations in the first quarter of the year. 3-room flats has an even distribution throughout and 4-room flats observe minor fluctuations over 2022.
Code 
sg_re_price_flat <- sg_re_price %>%
  filter(flat_type %in% c("3 ROOM", "4 ROOM", "5 ROOM"))

ggplot(data = sg_re_price_flat, aes(x = priceK, y = flat_type, fill = after_stat(x))) +
  
  geom_density_ridges_gradient(scale = 3, rel_min_height = 0.01) +
  
  theme_minimal() +
  
    labs(title = 'Resale Prices by Flat Type: {frame_time}',
       y = "Resale Price in Thousands (SGD)",
       x = "Flat Type") +
  
  theme(legend.position="none",
  text = element_text(family = "Garamond"),
  plot.title = element_text(face = "bold", size = 12),
  
  axis.title.x = element_text(size = 10, hjust = 1),
  axis.title.y = element_text(size = 10),
  axis.text = element_text(size = 8)) +
  
  scale_fill_viridis(name = "priceK", option = "D") +

  transition_time(sg_re_price_flat$Year) +
  ease_aes('linear')

Code 
ggplot(data = sg_re_price_2022_v1, aes(x = priceK, y = flat_type, fill = after_stat(x))) +
  
  geom_density_ridges_gradient(scale = 3, rel_min_height = 0.01) +
  
  theme_minimal() +
  
    labs(title = 'Resale Prices by Flat Type in 2022, Month: {frame_time}',
       x = "Flat Type",
       y = "Resale Price in Thousands (SGD)") +
  
  theme(legend.position="none",
  text = element_text(family = "Garamond"),
  plot.title = element_text(face = "bold", size = 12),
  
  axis.title.x = element_text(size = 10, hjust = 1),
  axis.title.y = element_text(size = 10),
  axis.text = element_text(size = 8)) +
  
  scale_fill_viridis(name = "priceK", option = "D") +
  
  transition_time(sg_re_price_2022_v1$Month) +
  ease_aes('linear')

4.1.2 Proportion of flat types in Singapore

The data clearly shows that 4-room apartments are the most popular in Singapore, accounting for nearly 50% of the flat types. 5-room and 3-rooms have relatively the same proportion of flats in Singapore - accounting for a quarter of each!

Code 
HDB_count  <- sg_re_price_2022_v1 %>%
  group_by(flat_type) %>%
  summarize(
    count = n()) %>% 
  mutate(hdb_pie_pct = round(count/sum(count)*100)) %>% 
  mutate(ypos_p = rev(cumsum(rev(hdb_pie_pct))),
         pos_p = hdb_pie_pct/2 + lead(ypos_p,1),
         pos_p = if_else(is.na(pos_p), hdb_pie_pct/2, pos_p))

ggplot(HDB_count, aes(x = "" , y = hdb_pie_pct, fill = fct_inorder(flat_type))) +
  geom_col(width = 1, color = 1) +
  coord_polar(theta = "y") +
  scale_fill_brewer(palette = "YlGnBu") +
  geom_label_repel(data = HDB_count,
                   aes(y = pos_p, label = paste0(hdb_pie_pct, "%")),
                   size = 4.5, nudge_x = 1, color = c(1, 1, 1), show.legend = FALSE) +
  guides(fill = guide_legend(title = "Flat Type")) +
  labs(title = "Proportion of Flat Types in Singapore")+
theme(legend.position = "bottom")+
theme_minimal()

4.1.3 Boxplot of Resale price by flat types

In descriptive statistics, a box plot or boxplot is a type of chart often used for EDA. Box plots visually show the distribution of numerical data and skewness through displaying the data quartiles (or percentiles) and averages.

When the median line appears in the middle of the boxplot, the data is normally distributed. We observe that:

  1. As expected, the median is the highest for 5-room flats, followed by 4-room and lastly 3-room flats.
  2. The dispersion (interquartile range) is the smallest for 3-room flats and largest for 5 room flats. This suggests that the 3 room flats are spread out around the median in comparison to 4 or 5 room flats.
  3. There is higher presence of outliers for 3-room flats, some even going up to 930000 (3rd quartile for 5-room flat)
Code 
t <- list(
  family = "Garamond",
  size = 19,
  face = "bold")

t1 <- list(
  family = "Garamond",
  size = 15,
  face = "bold")

fig <- plot_ly(
  data = sg_re_price_2022_v1,
  y = ~priceK,
  type = "box",
  color = ~flat_type,
  colors = "YlGnBu",
  showlegend = FALSE,
  boxmean = TRUE
) %>% 
  layout(title= list(text = "Boxplot of resale price in Thousands by flat type",font = t1),
         xaxis = list(title = list(text ='Flat Type', font = t1)),
         yaxis = list(title = list(text ='Resale Price in Thousands (SGD)', font = t1)))

fig

4.1.4 Is there an increase in resale price monthly by flat type?

Media outlets have been reporting that there has been a dramatic increase in the resale prices for HDBs. Real estate data is known to have many outliers, hence the best metric for comparison over time and distribution is median (they are unaffected by outliers). In the graph below, we observe that there is no discernible difference in the median prices, monthly, for 3 room, 4 room, 5 room flats,

Code 
ggplot(na.omit(sg_re_price_2022_v1),aes(x = flat_type, y = priceK)) +
  
  geom_boxplot(aes(fill = as.factor(Month)), color = "#c0c0c0") +
  
  stat_summary(fun.y = "mean", geom = "point", color = "black") +
  
  theme_minimal() +
  
  scale_fill_brewer(palette = "Paired") +

 labs( title = "Monthly Resale prices by Flat Types",
       y = "Resale price (1000s)",
       x = "Flat Type",
       fill = "Month") +

theme(text = element_text(family = "Garamond"),
        plot.title = element_text(hjust = 0.5, size = 12, face = 'bold'),
        plot.margin = margin(20, 20, 20, 20),
        legend.position = "bottom",
        
        axis.text = element_text(size = 10, face = "bold"),
        axis.title.x = element_text(hjust = 0.5, size = 12, face = "bold"),
        axis.title.y = element_text(hjust = 0.5, size = 12, face = "bold"))

Code 
na.omit(sg_re_price_2022_v1) %>% 
  filter(flat_type == "3 ROOM") %>% 
  ggplot(aes(x = flat_type, y = priceK)) +
  
  geom_boxplot(aes(fill = as.factor(Month)), color = "#c0c0c0") +
  
  stat_summary(fun.y = "mean", geom = "point", color = "black") +
  
  theme_minimal() +
  
  scale_fill_brewer(palette = "Paired") +

 labs( title = "Monthly Resale prices by Flat Types",
       y = "Resale price (1000s)",
       x = "Flat Type",
       fill = "Month") +

theme(text = element_text(family = "Garamond"),
        plot.title = element_text(hjust = 0.5, size = 12, face = 'bold'),
        plot.margin = margin(20, 20, 20, 20),
        legend.position = "bottom",
        
        axis.text = element_text(size = 10, face = "bold"),
        axis.title.x = element_text(hjust = 0.5, size = 12, face = "bold"),
        axis.title.y = element_text(hjust = 0.5, size = 12, face = "bold"))

Code 
na.omit(sg_re_price_2022_v1) %>% 
  filter(flat_type == "4 ROOM") %>% 
  ggplot(aes(x = flat_type, y = priceK)) +
  
  geom_boxplot(aes(fill = as.factor(Month)), color = "#c0c0c0") +
  
  stat_summary(fun.y = "mean", geom = "point", color = "black") +
  
  theme_minimal() +
  
  scale_fill_brewer(palette = "Paired") +

 labs( title = "Monthly Resale prices by Flat Types",
       y = "Resale price (1000s)",
       x = "Flat Type",
       fill = "Month") +

theme(text = element_text(family = "Garamond"),
        plot.title = element_text(hjust = 0.5, size = 12, face = 'bold'),
        plot.margin = margin(20, 20, 20, 20),
        legend.position = "bottom",
        
        axis.text = element_text(size = 10, face = "bold"),
        axis.title.x = element_text(hjust = 0.5, size = 12, face = "bold"),
        axis.title.y = element_text(hjust = 0.5, size = 12, face = "bold"))

Code 
na.omit(sg_re_price_2022_v1) %>% 
  filter(flat_type == "5 ROOM") %>% 
  ggplot(aes(x = flat_type, y = priceK)) +
  
  geom_boxplot(aes(fill = as.factor(Month)), color = "#c0c0c0") +
  
  stat_summary(fun.y = "mean", geom = "point", color = "black") +
  
  theme_minimal() +
  
  scale_fill_brewer(palette = "Paired") +

 labs( title = "Monthly Resale prices by Flat Types",
       y = "Resale price (1000s)",
       x = "Flat Type",
       fill = "Month") +

theme(text = element_text(family = "Garamond"),
        plot.title = element_text(hjust = 0.5, size = 12, face = 'bold'),
        plot.margin = margin(20, 20, 20, 20),
        legend.position = "bottom",
        
        axis.text = element_text(size = 10, face = "bold"),
        axis.title.x = element_text(hjust = 0.5, size = 12, face = "bold"),
        axis.title.y = element_text(hjust = 0.5, size = 12, face = "bold"))

4.2 Planning Area

4.2.1 Resale price by planning area over time

The ridge plot showcases:

  1. The resale prices across planning areas are more or less contained from 2017 to 2020. There is a marked increase (sharp right movement of all curves) in prices for all planning areas in 2020, then a gradual increase is observed from 2021.
  2. The distribution across planning areas is more or less constant till 2020. Post 2020, most of the distributions are flattening out indicating that the prices are widely distributed over these time periods. Bukit Timah, Central Area, Queenstown have 2 - 3 humps in their distribution implying that they have properties in the lower, medium and upper range.
Code 
ggplot(data = sg_re_price, aes(x = priceK, y = town, fill = after_stat(x))) +
  
  geom_density_ridges_gradient(scale = 3, rel_min_height = 0.01) +
  
  theme_minimal() +
    
  labs(title = 'Resale Prices by Planning Area: {frame_time}') +
  transition_time(sg_re_price$Year) +
  
  theme(legend.position="none",
  text = element_text(family = "Garamond"),
  plot.title = element_text(face = "bold", size = 12),
  
  axis.title.x = element_text(size = 10, hjust = 1),
  axis.title.y = element_text(size = 10, angle = 360),
  axis.text = element_text(size = 8)) +
  
  scale_fill_viridis(name = "priceK", option = "D") +

  ease_aes('linear')

Code 
ggplot(data = sg_re_price_2022_v1, aes(x = priceK, y = town, fill = after_stat(x))) +
  
  geom_density_ridges_gradient(scale = 3, rel_min_height = 0.01) +
  
  theme_minimal() +
  
  labs(title = 'Resale Prices by Planning Area in 2022, Month: {frame_time}') +
  
  theme(legend.position="none",
  text = element_text(family = "Garamond"),
  plot.title = element_text(face = "bold", size = 12),
  
  axis.title.x = element_text(size = 10, hjust = 1),
  axis.title.y = element_text(size = 10, angle = 360),
  axis.text = element_text(size = 8)) +
  
  scale_fill_viridis(name = "priceK", option = "D") +
  
  transition_time(sg_re_price_2022_v1$Month) +
  ease_aes('linear')

4.3 Flat type comparison by planning area

4.3.1 Proportion of Flat types by Planning area in Singapore

We now look into the proportion of flat types by planning areas. We observe that:

  1. Toa Payoh, Ang Mo Kio, Bedok, Central Area have the highest proprtion of 3-room flats, whereas Bukit Timah and Paris Ris have the least.
  2. Yishun, Woodlands, Serangoon have the highest proportion of 4-room flats.
  3. Sengkang, Punggol have the highest proportion of 5-room flats.
Code 
prop_sg <- sg_re_price_2022_v1%>%
group_by(town, flat_type) %>%
summarize(ft_count = n())%>%
mutate(ft_pct = scales::percent(ft_count/sum(ft_count)))

prop <- ggplot(prop_sg, 
                aes(town, ft_count, fill = flat_type)) + 
  geom_bar(stat="identity") +

  labs(title = "Proportion of Flat types by Planning area in Singapore", x = "Planning Area", y = "Count", fill = "Flat Type") +
  
  theme_minimal() +
  
  scale_fill_viridis(discrete = T, option = "E") +
  
  theme(text = element_text(family = "Garamond"),
        plot.title = element_text(hjust = 0.4, size = 15, face = 'bold'),
        plot.margin = margin(20, 20, 20, 20),
        legend.position = "bottom",
        axis.text = element_text(size = 8, face = "bold"),
        axis.text.x = element_text(angle = 90, vjust = 0.5, hjust=1),
        axis.title.x = element_text(hjust = 0.5, size = 12, face = "bold"),
        axis.title.y = element_text(hjust = 0.5, size = 12, face = "bold"))

ggplotly(prop)
Code 
prop_sg <- sg_re_price_2022_v1%>%
group_by(town, flat_type) %>%
summarize(ft_count = n())%>%
mutate(ft_pct = scales::percent(ft_count/sum(ft_count)))

prop_sg
# A tibble: 78 × 4
# Groups:   town [26]
   town        flat_type ft_count ft_pct
   <chr>       <chr>        <int> <chr> 
 1 ANG MO KIO  3 ROOM         524 56%   
 2 ANG MO KIO  4 ROOM         287 30%   
 3 ANG MO KIO  5 ROOM         132 14%   
 4 BEDOK       3 ROOM         555 43.7% 
 5 BEDOK       4 ROOM         462 36.4% 
 6 BEDOK       5 ROOM         253 19.9% 
 7 BISHAN      3 ROOM          65 17%   
 8 BISHAN      4 ROOM         200 51%   
 9 BISHAN      5 ROOM         124 32%   
10 BUKIT BATOK 3 ROOM         281 35%   
# … with 68 more rows

4.3.2 Resale prices of different flat types by planning area

The dumbell chart showcases the minimum and maximum prices for each planning area.

  1. Woodlands, Sembawang, Choa Chu Kang provide more affordable flats with their maximum price being below SGD 800,000. Conversely, most of the other regions have wider distribution of prices ranging from mid to exorbitant.
Code 
areas = unique(sg_re_price_2022_v1$town)
PA = c()
min_price = c()
max_price = c()

for(area in areas){
  df <- sg_re_price_2022_v1[sg_re_price_2022_v1$town == area,]
  PA = c(PA, area)
  min_price = c(min_price, min(df$priceK))
  max_price = c(max_price, max(df$priceK))
}

df <- data.frame("Planning_Area" = PA, "Min_Price" = min_price, "Max_Price" = max_price, check.names=FALSE)
Code 
dumbell <- plot_ly(df) %>% 
 add_segments(x = ~Min_Price, xend = ~Max_Price, y = ~Planning_Area, yend = ~Planning_Area, showlegend = FALSE) %>% 
 add_markers(x = ~Min_Price, y = ~Planning_Area, name = "Min", color = I("#e66819")) %>% 
 add_markers(x = ~Max_Price, y = ~Planning_Area, name = "Max", color = I("#bf0d31")) %>% 
 layout(
    title = "Resale Price in 1000s Difference Across Planning Areas",
    xaxis = list(title = "Resale Price in 1000s (SGD)"),
    yaxis = list(title = "Planning Area"),
    margin = list(l = 70)
  )

dumbell

The dumbell chart is a parochial view of the range of resale price. Hence, we further delve into the box and whiskers plot of 3-room, 4-room, and 5-room flat types across the planning regions.

We see that areas such as Jurong West and Tampines have the widest price range for 5-room type flats, especially towards the lucrative side of the price. In Marine Parade, the 5-room flat types were significantly higher priced than 3-room and 4-room type flats. The opposite was true for areas such as Serangoon.

Code 
#have to change theme and everything else

p <- ggplot(data = sg_re_price_2022_v1, aes(x = town, y = priceK)) +
  
  geom_boxplot(aes(fill = as.factor(flat_type))) +
  
  theme_minimal() +
  
  scale_fill_brewer(palette = "Dark2") +

 labs( title = "Resale prices in Thousands in 2022 by planning area",
       fill = "Flat Type",
       y = "Resale price (1000s)",
       x = "Planning Area") +

theme(text = element_text(family = "Garamond"),
        plot.title = element_text(hjust = 0.2, size = 12, face = 'bold'),
        plot.margin = margin(20, 20, 20, 20),
        legend.position = "bottom",
        
        axis.text = element_text(size = 8, face = "bold"),
        axis.text.x = element_text(angle = 90, vjust = 0.5, hjust=1),
        axis.title.x = element_text(hjust = 0.5, size = 12, face = "bold", color = "#899499"),
        axis.title.y = element_text(hjust = 0.5, size = 12, face = "bold", color = "#899499"))

ggplotly(p)

4.2 Area & Age by Flat Type

4.2.1 Correlation for Area, Age with Resale Price

The correlation chart shows that there is a positive relationship between area of property and resale price i.e. larger the property size gets, higher the resale price. Resale price and age have a weak negative relationship.

Code 
ggscatterstats(
 data = sg_re_price_2022_v1,
x = floor_area_sqm,
y = priceK,
marginal = FALSE) +
  
  theme_minimal() +
  
  labs(title = 'Correlation of resale price (in 000s) and floor area (sqm)', x = "Floor Area", y = "Resale Price in 000s") +

theme(text = element_text(family = "Garamond"),
        plot.title = element_text(hjust = 0.2, size = 15, face = 'bold'),
        plot.margin = margin(20, 20, 20, 20),
        legend.position = "bottom",
        
        axis.text = element_text(size = 10, face = "bold"),
        axis.title = element_text(size = 12, face = "bold"))

Code 
ggscatterstats(
 data = sg_re_price_2022_v1,
x = property_age,
y = priceK,
marginal = FALSE) +
  
  theme_minimal() +
  
  labs(title = 'Correlation of resale price (in 000s) and floor area (sqm)', x = "Floor Area", y = "Resale Price in 000s") +

theme(text = element_text(family = "Garamond"),
        plot.title = element_text(hjust = 0.2, size = 15, face = 'bold'),
        plot.margin = margin(20, 20, 20, 20),
        legend.position = "bottom",
        
        axis.text = element_text(size = 10, face = "bold"),
        axis.title = element_text(size = 12, face = "bold"))

4.2.2 Treemap for Age and Area

Through the treemap, we see on overall glimpse of flat types by prices per square meter and resale prices in 000s across all planning areas.

Code 
treemap_area <- treemap (sg_re_price_2022_v1,
        index= c("flat_type", "town"),
        vSize= "floor_area_sqm",
        vColor = "priceK",
        type="manual",
        palette = mako(5),
        border.col = c("black", "white"),
        title="Properties for resale",
        title.legend = "Median Price (1000s)"
        )

Code 
treemap_age <- treemap (sg_re_price_2022_v1,
        index= c("flat_type", "town"),
        vSize= "property_age",
        vColor = "priceK",
        type="manual",
        palette = "Blues",
        title="Properties for resale",
        title.legend = "Median Price (1000s)"
        )

5. References

  1. Dag, Osman, ‘How to remove Outliers from data in R?’, ‘Data Science Team’, March 4, 2022 (link
  2. White, Kerber, ‘How to Draw a plotly Boxplot in R (Example)’, ‘Statistics Globe’, February, 2022 (link)
  3. ‘Setting the Font, Title, Legend Entries, and Axis Titles in R’ (link)
  4. ‘R: ggplot stacked bar chart with counts on y axis but percentage as label’, 2017 (link)
  5. ‘Ridgeline Plots’, Github (link)
  6. McLeod, Saul, ‘What does a boxplot tell you?’, ‘Simply Psychology’, 2019 (link)
  7. Wenyi, Wang, ‘An analysis of Singapore property resale market price based on transaction from 2017 to 2019’, July 17, 2020 (link)
  8. Qingquan, XIA, ‘Assignment 5’, ‘RPubs’, August, 2020 (link)
  9. TS Kam, ‘Hands-on Exercise 8: Treemap Visualisation with R’, July, 2019, (link)