Interactive Learning Roadmap - Part 2

Visualisation Requirements

Now that we understand the data better, and explored different layouts, what do we need in our final publish visualisation?

• Skills and skill groupings are relational so we need these as objects,
• Skills are connected objects, so we will need line connections,
• There is some directionality between objects,
• However skills and skill set groupings are not all sequential, and the groupings may need to be spliced together,
• Skills, tips and groupings need to be identifiable with labels/ text,
• Experience and level needs to be identifiable but since they are numeric these can be represented by text, shapes or colours,
• The output needs to be scalable as more data is added.

Let’s see if network graphs meet these requirements.

Network Graphs

Although the data is actually connected and seems relational, the relationships between each dataframe is not currently captured. According to the tidygraph release notes:

There’s a discrepancy between relational data and the tidy data idea, in that relational data cannot in any meaningful way be encoded as a single tidy data frame. On the other hand, both node and edge data by itself fits very well within the tidy concept as each node and edge is, in a sense, a single observation. Thus, a close approximation of tidyness for relational data is two tidy data frames, one describing the node data and one describing the edge data.

Static Network Graph

Let’s now create the two dataframes for the node data and the edge ( or line links) data following this great network analysis with R tutorial.

# Add id to setupR
setupR <- setupR %>% 
      rowid_to_column("id")
# In the line plots we used the dots for the skills, so create nodes_1 for each of the skills with unique ids, and the node name as label
(nodes_1 <- tibble(label=setupR$skill) %>% 
      # Adds a column with the values from the row ids at the start of the data frame
      rowid_to_column("id"))

## # A tibble: 9 x 2
##      id label             
##   <int> <chr>             
## 1     1 Install RStudio   
## 2     2 Install R         
## 3     3 Install R packages
## 4     4 Load Packages     
## 5     5 Update RStudio    
## 6     6 Update R          
## 7     7 Update R packages 
## 8     8 Basic debugging   
## 9     9 Browser debugging

# Since this dataframe has a sequential route, let's create a route table with from and to ids - this code could be improved later)
route <- tibble(from = setupR$id,
                to = c(setupR$id[-1],setupR$id[9]))

(edges_1 <- route %>%  
            # Use a left join where the key from in route and id in setupR
            left_join(setupR, by = c("from" = "id")) %>%
            select(from,to) %>% 
            # If the edges have magnitude then this is referred to as weight. Initially set the weight as default of 1
            mutate(weight=1))

## # A tibble: 9 x 3
##    from    to weight
##   <int> <int>  <dbl>
## 1     1     2      1
## 2     2     3      1
## 3     3     4      1
## 4     4     5      1
## 5     5     6      1
## 6     6     7      1
## 7     7     8      1
## 8     8     9      1
## 9     9     9      1

# Skip along the tutorial to the tidygraph and ggraph sections, to create a network object using tbl_graph
(routes_tidy <- tbl_graph(nodes = nodes_1, edges = edges_1, directed = TRUE))

## # A tbl_graph: 9 nodes and 9 edges
## #
## # A directed multigraph with 1 component
## #
## # Node Data: 9 x 2 (active)
##      id label             
##   <int> <chr>             
## 1     1 Install RStudio   
## 2     2 Install R         
## 3     3 Install R packages
## 4     4 Load Packages     
## 5     5 Update RStudio    
## 6     6 Update R          
## # ... with 3 more rows
## #
## # Edge Data: 9 x 3
##    from    to weight
##   <int> <int>  <dbl>
## 1     1     2      1
## 2     2     3      1
## 3     3     4      1
## # ... with 6 more rows

# Take a look the class of this routes_tidy object
class(routes_tidy)

## [1] "tbl_graph" "igraph"

# Set seed for reproduceability of graph
set.seed(123)
# Now use the ggraph function to plot the network object
ggraph(routes_tidy) + 
      geom_edge_link() +
      geom_node_text(aes(label = label)) + 
      geom_node_point() + 
      theme_graph()

## Using `nicely` as default layout

A network graph has the functionality to directly label objects,

This network graph is static, and not quite ready for publishing.

Interactive Network Graph

Let’s look to customise further and create an interactive version the network graph.

Let’s try the visNetwork R package to plot an interactive network. Although it is an interface to the ‘vis.js’ JavaScript library, it is compatible with the tidygraph tidyverse format and syntax.

# As per the documentation, nodes must be a dataframe with least one column id, and edges must be a dataframe with at least from and to columns for visNetwork. Plot using the default options
visNetwork(nodes_1,edges_1) %>% 
     # The direction of the edges is meaningful so this is a directed network and include arrows
      visEdges(arrows = "middle") %>%
      # Set seed for reproduceability of graph
      visLayout(randomSeed = 123)

Shapes

This network graph uses solid and coloured shapes rather than outlines or small dots as these are easier to perceive. We can also use a different shape for the start of the journey to differentiate the objects and give the reader a starting reference point.

Colour Scales

We can use colour to distinguish groups, represent values or to highlight. For these groups we need to represent the groups so use a qualitative palette. We can look to highlight objects with colour on interactions such as clicks.

# Define the qualitative colour scale to represent 9 distinctive skill groups using colorbrewer http://colorbrewer2.org/#type=qualitative&scheme=Set1&n=9
pal <-  RColorBrewer::brewer.pal(9, "Set1")

Next add some attributes to expanded nodes and edges dataframes with all the skill set groupings.

# Create the nodes dataframe with the colour group and additional attributes. This code code be improved with a function later to avoid the repetition.
nodes <- bind_rows(setupR %>%
                          select(-id) %>% 
                          mutate(color=rep(pal[1], dim(setupR)[1])) %>%
                          mutate(group=rep("setupR",dim(setupR)[1])),
                    objectsR %>%  
                         mutate(color=rep(pal[2], dim(objectsR)[1]))%>%
                         mutate(group=rep("objectsR", dim(objectsR)[1])),
                    importR %>% 
                         mutate(color=rep(pal[3], dim(importR)[1]))%>%
                         mutate(group=rep("importR", dim(importR)[1])),
                    functionsR %>% 
                         mutate(color=rep(pal[4], dim(functionsR)[1]))%>%
                         mutate(group=rep("functionsR", dim(functionsR)[1])),
                    datamungingR %>% 
                         mutate(color=rep(pal[5], dim(datamungingR)[1])) %>%
                         mutate(group=rep("datamungingR", dim(datamungingR)[1])),
                    vizR %>% 
                         mutate(color=rep(pal[6], dim(vizR)[1]))%>%
                         mutate(group=rep("vizR", dim(vizR)[1])),
                    publishR %>%
                         mutate(color=rep(pal[7], dim(publishR)[1])) %>%
                         mutate(group=rep("publishR",dim(publishR)[1])),
                    statsR %>%
                         mutate(color=rep(pal[8], dim(statsR)[1])) %>%
                         mutate(group=rep("statsR",dim(statsR)[1]))
                    )  %>% 
            rename(label=skill) %>% 
            # Setting the shape as circle means the labels are inside the circle and the value doesn't work. I would like to keep the labels outside the dot https://stackoverflow.com/questions/39674927/how-have-labels-inside-the-scaled-nodes-in-visnetwork
            mutate(shape="dot") %>% 
            rename(value=experience) %>% 
            # Add a column with the values from the row ids at the start of the data frame
            rowid_to_column("id")
# Print using glimpse to view nodes dataframe
nodes %>%  
      glimpse()

## Observations: 53
## Variables: 9
## $ id    <int> 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 1...
## $ label <chr> "Install RStudio", "Install R", "Install R packages", "L...
## $ value <dbl> 5, 5, 5, 5, 5, 3, 5, 3, 1, 5, 5, 5, 5, 5, 5, 5, 5, 3, 3,...
## $ level <dbl> 1, 1, 1, 1, 1, 1, 3, 1, 3, 1, 1, 1, 1, 1, 1, 1, 1, 3, 3,...
## $ tip   <chr> "Download from RStudio website", "", "CRAN install.packa...
## $ date  <chr> "Apr-17", "Apr-17", "Apr-17", "Apr-17", "Apr-17", "Apr-1...
## $ color <chr> "#E41A1C", "#E41A1C", "#E41A1C", "#E41A1C", "#E41A1C", "...
## $ group <chr> "setupR", "setupR", "setupR", "setupR", "setupR", "setup...
## $ shape <chr> "dot", "dot", "dot", "dot", "dot", "dot", "dot", "dot", ...

# Change the first start node to a square to differentiate the start object from the other skill objects. Use the column name in case the ordering of columns changes
nodes$shape[1] <- "square"
# Create the edges with the from and to numeric values
edges <- tibble(from = nodes$id) %>%
       # to = c(nodes$id[-1],nodes$id[dim(nodes)[1]])) #old code
      # Use lead function from dplyr to add column with the next values https://stackoverflow.com/questions/33038240/add-row-value-to-previous-row-value-in-r
      mutate(to=lead(from)) %>% 
      # Add default weight variable of 1
      mutate(weight=1)
# Update the ends of groupchains so that the 'to' from value is NA. This has been updated using a logical vector and dplyr lead function
edges$to <- ifelse(nodes$group==lead(nodes$group),
                       edges$to,
                       NA)
# Add splices ie edges for nodes that are interconnected using merge
splices<- data.frame(splicefrom=c("Load Packages", "Load Packages", "Dataframes", "Dataframes", "Dataframes","Dataframes","Environments","Matrices","Environments","Lists","Vectors"),
                     spliceto=c("Object assignment", "R Markdown", "Numerical Indexing", "Summary statistics", "Base plots", "txt", "Create Functions","Reshape","Scoping rules","Functional programming","Vector functions"),
                     stringsAsFactors = FALSE) %>% 
      merge(.,nodes %>%  select(id,label), by.x="splicefrom", by.y="label") %>% 
      rename(from=id) %>% 
      merge(.,nodes %>%  select(id,label), by.x="spliceto", by.y="label") %>% 
      rename(to=id) %>% 
      mutate(weight=1) %>% 
      select(-splicefrom,-spliceto)
# Now bind the spliced edges to the edges dataframe
edges <- bind_rows(edges,
                   splices)

Now plot an interactive visNetwork again following the very good visNetwork vignette.

# Nodes data.frame for legend
lnodes <- data.frame(label = c("Start", "Beginner", "Intermediate","Experienced"),
                     shape = c( "square", "dot","dot","dot"), 
                     value= c(5,1,3,5),
                     title = "Skill Experience",  
                     # Color the non start dots as light grey as this is the non highlight colour
                     color=c("red", "lightgrey", "lightgrey", "lightgrey"),
                     id = 1:4)
# Use the visNetwork function again with value ie the experience as the size of the node
visNetwork(nodes %>%
                 # Concatenate the title for the tooltip
                 mutate(title=paste0("Skill: ",label,
                                     "<br>Skill Group: ",group, 
                                     "<br>Tips: ",tip))  ,
           edges,
           main =("Interactive Learning Journey Roadmap with R"),
           submain= ("Use mouse to zoom")) %>% 
           # Now that we have a legend we can update the submain as the start should be intuitive
           # submain="Click on the square to start the journey") 
      visGroups(groupname = group)  %>% 
      # The direction of the edges is meaningful so for this directed network we will include arrows
      visEdges(arrows = "middle") %>% 
      # Create legend for shapes
      visLegend(addNodes = lnodes,
                useGroups=FALSE,
                position="right",
                main="Skill Experience Categories") %>% 
      # Add a grouping drop down menu and highlight the nodes on selection
      visOptions(selectedBy = "group",
             highlightNearest = TRUE, 
             nodesIdSelection = FALSE) %>%
      # Set seed for reproduceability of graph
      visLayout(randomSeed = 123)

Lastly, as variations to the visualisation, we could change the filter menus to highlight the graph by skill level or to make it scalable, edit the graph using the function visOptions(manipulation = TRUE) or use skill function created in Part 1.