This essay has been submitted by a student. This is not an example of the work written by our professional essay writers.
Visualization emerged as a major technology with multiple interesting applications such as social networks, information retrieval, biochemical pathways and word dictionaries. As data become very large nowadays, visualization plays a major part in social network analysis that reveals the hidden pattern in data and helps users to understand the connections between data elements (humans, communities and companies) and their internal relationships. Due to its numerous challenges, recent developed applications with new capabilities contributed to facilitate the exploration and analysis for visualization tasks. In this survey, we give background knowledge about different visualization graph layouts, their applications and user navigation and exploration techniques. Then, we present other visualization layouts. Finally, we give conclusion and point out some problems for future research.
In the past six years, online social networks such as (Facebook, MySpace, Hi5 and tagged) are lunched as one of the dominant web services in the internet and with the growing number of users in online social networks, new datasets are becoming available as users spend a lot of time to connect with friends and participate in the internet growth through the interactions between them in the form of electronic materials such as the published blogs, uploaded movies, pictures, and resumes on YouTube, Flickr, and LinkedIn websites.
Understanding social networks are difficult with the existing tools like Facebook's TouchGraph  or LinkedIn's MySpace. They are not sophisticated enough to analyze such communities like these to discover the social structure and hierarchy gaps. These tools do not give a deep insight on the network, just a simple overview of the network from the user perspective without an overall overview of the entire network with deep network understanding.
Analysts focus on the relations between the elements of network rather than the characteristics of the element itself, e.g., professor has characteristics like (name, affiliation, departmentâ€¦etc), the same as student (course, GPAâ€¦etc). To visualize such a network of professors and students, the analyst is concerned with the relationship between students and professors not with their characteristics. So, he will be able to reveal the hidden pattern in the network structure based on the relationships. But, the analyst doesn't want to perform analysis on data based on fixed (static) models according to previous assumptions on these models, he wants to explore data dynamically (dynamic modeling) without the need for priori assumptions about the nature of the data.
Here we are explicitly interested in social network analysis with its applications that capable of visualizing large amount of complex data and how it can handle them nicely with automatic procedures of powerful mining techniques that can extract high level non-trivial information from large repositories.
The paper is organized as follows: we introduce social networks graph layouts (node-link diagram and matrix-based representations), their applications and with some future research directions. After that we show other different s graph network layouts. Finally, Conclusion and future work are provided.
SOCIAL NETWORK GRAPH LAYOUTS
Social network is consisted of group of actors linked to each other through relationships. Typical examples of networks are networks of individuals (people in a specific culture, people in communities, people in similar working contexts, etc.) or networks of organizations (companies, cities, political parties). The basic assumption behind it is that a large variety of real world phenomena could be explained and described in terms of relationships between actors. As show in Fig. 1, the actors can be connected by a link that represents a relation: some sort of acquaintance or social relationship (e.g., work, high school, camp).
There are several types of social network representations but the most popular ones are: 1) node-link diagrams, and 2) matrix-based representation.
Node-link diagrams can be considered as the first kind of drawing ever used to represent relations between objects. The basic graph layout problem is very simple. Given a set of nodes with a set of relations (edges), it only needs to calculate the positions of the nodes and draw each edge as curve. However, to make graphical layouts understandable and useful is very hard. Basically there are generally accepted aesthetic rules [23, 24], which include:
Distribute nodes and edges evenly.
Avoid edge crossing.
Display isomorphic substructures in the same manner.
Minimize the bends along the edges.
Figure. 1 Example for social network from source 
Though these aesthetic rules are generally accepted, they are not equally important. However, most of the time, it is quite impossible to meet all rules at the same time. Some of them conflict with each other. Some of them are very computationally expensive. Thus, practical graphical layouts are usually the results of compromise among the aesthetics.
The kind of patterns an analyst looks for is either one of these two types: social communities or social roles. Every measure or methodology is always oriented towards one of these two kinds. When searching for social communities, the analyst tries to identify the global structure of the network, searching for the hidden patterns, that is, if interesting (unexpected) conglomerates appears, and if these conglomerates can be explained in terms of a shared quality or event. When looking at social roles, instead, the analyst is concerned with the position of single entities, trying to find those nodes that have great importance in the network structure. So, analyst focuses on two set of features in node-link diagram 1) global features, and 2) local features
The most well known global features are 1) global clustering coefficient; 2) average path length, and 3) degree distribution.
The global clustering coefficient is a measure of degree to which nodes in a graph tend to cluster together. Evidence suggests that in most real-world networks, and in particular social networks, nodes tend to create tightly knit groups characterized by a relatively high density of ties , it is defined by the following equations:-
In equation 1, represents the number of neighbor nodes. While, equation 2 is defined as the average clustering coefficient of its node for some graph and indicates the number of nodes in the graph.
The average path length is defined as the average number of steps along the shortest paths for all possible pairs of network , it is described by the following equation:-
Where d is the distance between two nodes u and v in Graph G, without loss of generality, it measures how distant, on average is a path from a node to any other node in the graph.
The degree of a node in a network is the number of connections it has to other nodes and the degree distribution is the probability distribution of these degrees over the whole network , it is defined by the following equation:-
Where each of n nodes is connected (or not) with independent probability p (or 1 âˆ’ p). These measures are used to capture global information for an initial analysis and are useful to compare different networks to one another.
It is important to highlight that there exist computational methodologies that detect and give the option to draw clustered subgroups. In fact, if the above features allow numerical notations to be described on graph groupings, it is also important to give attention to techniques that make these groups available on the screen. Spring embedding is a drawing methodology where it group nodes that are highly connected in certain positions on a plan. Other common computational methodologies are factor analysis and multidimensional scaling (MDS) that analyze the attributes of nodes and project them on a 2D plane so that similar nodes fall next to one another.
If we are concerned with reveal the structure between the highly connected nodes, then analyst will look at their local features. One wants to calculate what is called the centrality of a node. The most famous types of centrality: 1) degree centrality; 2) betweenness Centrality; and 3) closeness centrality.
Degree centrality is defined as the number of links incident upon a node (i.e., the number of ties that a node has) more connections mean more relevance , it is defined by the following equation:-
Where deg(v) is the number of links node v has with other nodes.
Betweenness centrality discovers the nodes that are central and able to connect between clusters that would be disconnected in the absence of these nodes. In other terms it is a centrality measure of a vertex within a graph where vertices that occur on many shortest paths between other vertices have higher betweenness than those that do not , it is described by the following equation:-
Where is the number of shortest paths between s and t and is the number of these shortest paths which pass through v.
Closeness centrality mean shortest-path length, as it gives higher values to more central vertices, if a node is very close to a large number of nodes, it can monitor the overall flow of information and thus it has a high visibility on what happens in the network, it is defined by the following equation:-
Where is the distance between and is the number of edges to traverse in their shortest path.
After grouping nodes based on their global and local features, the nodes may form specific kinds of networks that are worth to mention small-world and scale-free networks. These are special kinds of networks that are able to describe a large set of real-world phenomena
It is a type of network in which most nodes are not neighbors of one another, but most nodes can be reached from every other by a small number of hops or steps , when it has a high clustering coefficients and average shortest path at the same time. A network with such features contains highly clustered nodes, but still it is not difficult for an individual to reach any other node of the network a small. A small world network, where nodes represent people and edges connect people that know each other, captures the small world phenomenon of strangers being linked by a mutual acquaintance.
It is a type of network, whose degree distribution follows a power law, it has only few clustered nodes which are far more connected than the others, and these nodes act as central hubs. That is, the network has either few highly connected nodes or many poorly connected ones as shown in Fig. 2. Scale-free networks are noteworthy because many empirically observed networks appear to be scale-free, including the protein networks, citation networks, and some social networks 
Figure. 2 Scale-free network
Node-link diagram applications
The users in online social network applications construct massive graph structures of interconnecting nodes, where the relationships between the articulation points (central nodes connect between communities) are buried and can only be clear by browsing each node profile. So, capturing the patterns between communities is hard and this will lead to problems like the ability for members to explore their online communities. Vizster  is a rich tool that provides node-link diagram representation that targets online social networks, it solves the shortages in the existing applications; it aims to explore the relationships between actors in communities, and discover the connections between communities as shown in Fig. 3. Vizster provide a number of interactive navigation and exploration starting from visual search, analysis nodes based on actors attributes, customizing springembedded and clustering communities automatically as shown in Fig 4.
Vizster design an effective and efficient ways to map data to visual structures. It takes in consideration three points when build the network structure 1) visual objects to use and how to map them on data, 2) how to place objects on the screen (an interesting taxonomy is provided in ), 3)how to map retinal attributes (i.e., color, size, texture) to data features of interest as shown in Fig. 5.
The layout is computed in real-time that avoids interactive delays induced by layout computation and allows the user to participate in the layout process by interactively dragging nodes to tease apart communities
Allow users to expand visualized networks while maintaining landmarks.
Interactive highlighting that aid network understanding
Panning, zooming, and distortion techniques are provided
Interactive search and attribute visualization "X-Ray" as shown in Fig 6 and 7.
Not effective at showing the global structure such as central actors .
Not effective in visualizing larger networks .
Figure. 3 Vizster
Figure. 4 Vizster clusters communities automatically
Figure. 5 Vizster Attribute Visualization using "X-Ray" mode the number of friends
Figure. 6 Vizster X-ray mode gender visualization
Figure. 7 X-ray Mode visualizing genders, search hits, mouse-over
highlight and community structures
Social network  analysis presented as a prominent technique to reveal the unclear parts of relationships with the ability to discover the importance of connections in networks. However, interactive navigation of networks still challenging because: (1) it is hard to discover patterns and understand the structure of networks with many data vertices and edges, and (2) the lack of visual outputs that provide interactivity between the analysts and networks still poor and leave unanswered questions
SocialAction is another node-link diagram representation that helps users to view social networks based on attribute ranking shown in Fig. 6 and coordinate views as shown in Fig 7. Users can have better visualization by filtering objects and find their corresponding outliers, compressing several nodes in a single node link structure for better visualization, highlighting desired communities of interest.
Figure. 6 SocailAction Attribute ranking
Ranking social networks: SocialAction gives users the ability to rank nodes by their structural position by selecting a ranking of interest from a drop-down menu. When users choose a ranking, all of the nodes are ranked according to chosen criteria in the ordered list. Each ranking has a corresponding color, ranging from green to black to red, based on its value. This helps visualize each node's position among. all ranked entities
Filtering by Rankings: SocialAction also gives users the flexibility to limit the number of nodes in both the ordered list and the network view based on their rankings where users can ignore portions of the network that do not meet their criteria as shown in Fig 8.
Comparing Rankings with Scatterplots: A structural analyst is concerned with in the nodes that meet his interests across two rankings. SocialAction presents 2-D projection as a scatterplot. A scatterplot reveals the form direction and strength of a relationship between two features, in addition to identifying outliers easily. Users are given the options to select two features that form the axes for a scatterplot
Multiplex ranking: Nodes that do not have any links of the selected type are faded
SocialAction doesn't work effectively when the number of nodes increases (more than thousands).
Figure. 7 SocailAction coordinate views
Figure. 8 SocailAction with filtering by ranking
All previous online social networks can't scale well when the number of nodes increases dramatically from thousands to millions of nodes which will turn the overview graph into an all-black copy paper; and the target users of the system are generally ordinary people that will be frustrated.
HiMap  is an online social network visualization tool that able to view large scale networks effectively. It meets the following four important goals:-
Each graph view of the network should be adaptively visualized in a readable manner that is easy to be comprehended, independent with its scale, topology and the screen size to display.
A suite of navigation methods should be provided so that it is capable to visualize and diagnose every detail of the network.
Smooth animations should be presented between any view changes, so as to keep user's momentum
The visualization system should run fast enough and keep lightweight: it could catch up with the animation.
HiMap also is presented with Stable Kamada-Kawai layout algorithm for Clustered graphs (SKK-C) that works in a recursive manner and aims to avoid cluster overlapping. Also, a novel adaptive data loading method that works in the offline mode, described in three steps:-
Rank the visual items in the same hierarchy by pre-defined importance metric.
Quantify the number of displayed visual items according to the current screen size
Adaptively load the visual items by their rankings and the available visual item budget.
HiMap also includes a set of exploration and navigation tools to preserve the readability of the graph view such as selecting, dragging, customized zooming, panning and customized animations for each HiMap interactions upon view changes. HiMap pipeline is shown in Fig. 9
Figure. 9 HiMap pipeline
HiMap pipeline is divided into parts 1) the offline data manipulation which includes data collecting, cleaning, and hierarchical clustering stage. It prepares the graph data required for further on-line visualization of the target social networks; and 2) the online adaptive visualization involves data loading, graph layout, projection and rendering stages. Finally, we can summarize HiMap advantages in the following points
Offline data manipulation where data are stored in the database
Visual Metaphors where the visual items are rendered to facilitate the human perception
Adaptive data loading that maintain comfortable visual density and readability subject to the changing screen size.
Customized zooming operations where semantic zooming will bring new visual items to the view upon magnifying and retract some old visual items with low importance ranking upon minifying
In matrices, rows and columns refer to nodes (vertices) in node-link diagrams, while, cells refers to relationships (edges). Although, node-link diagrams are the most used and popular representations in social networks, they are good at showing the overall structure of the connections between communities, but, Ghoniem et al.  showed that density has a strong impact on readability in node-link diagrams. Focusing on basic readability tasks such as finding an actor or determining if two actors are linked, they conclude that node-link diagrams perform badly to reveal the connections with nodes for dense networks even with few (e.g. 20) nodes. Because node-link diagrams become unreadable in dense communities and around high-degree hub nodes, they do not lend themselves to community analysis . Fig. 10 shows an example for matrix representation. So, we summarize matrix features over node-link diagrams in the following points:
In matrix visualizations the objects cannot overlap; thus resolving the cluttering effect and improving readability
Rows and columns can be permuted according to optimization algorithms .
Matrices provide flexible ways to display attribute information like size, color and shape of objects.
Most of matrix-based representations and node-links diagrams are examples for such one single static view for the network. There are a set of interactive techniques to accomplish exploration and navigation tasks and allow seeing things from different point of views, expanding the user capabilities, and changing the way objects are displayed on the screen, before moving to mention matrix-based representation tools, we give brief description on interaction.
Interaction and navigation
Mantra presented the idea of navigation  in a descriptive sentence "overview first, zoon and filter details-on- demand" which capture the kind of activity the analyst wants to perform in exploration tasks. If the analyst wants to inspect a specific area of interest for more social network analysis, he will use a set of customized techniques to tune the research focus on the interesting parts, to filter out the unwanted information and to try visualizing network with different configuration to capture al the aspect of the underlying structure of the network.
Figure. 10 Matrix-based representations
Exploration and navigation techniques:-
Interactive highlight: this navigation feature is available in node-link diagrams, where the user is able to select specific number of connections and paths for specific node and can group more than one node to see the paths that connect them, making their visual appearance more clear and deemphasizing the rest. In this way the user can focus on the area of interest without being distracted by other paths and nodes.
Details-on-demand: when there is a need to access additional information on a single item or small groups that are not available on the screen, the user select one or more elements and ask for detailed information on them, e.g., attributes displayed in textual form either in the visualization or in a separate pane.
Filtering: it is a popular technique in a number of applications like NodeTrix , MatrixExplorer  and Vizster . The main idea is to filter out or limit the group of items displayed on the screen by selecting range of values, the user filter out items using check box, sliders, menus and queries. Filtration applied on the visualized network and displays only the items that meet the criteria specified by ranges. For example, a filtering tool may allow limiting the number of displayed nodes of a graph according to their attributes (number of links between nodes), these nodes that meet these criteria appear and others disappear where the value can be changed using slider to try different values of thresholds which is an interactive way of filtering.
Focus + context: the idea of focus plus context is that when the screen contains too many objects, the user wants to see the object of primary interest presented in full detail accepts a reduction of resolution for the remaining objects. While preserving an overview-impression of all the surrounding information
Interactive clustering: the user needs a way to reduce the number of elements displayed on the screen, and change the structure of the network, depending on the attribute or relation visualized. He can abstract from details and observe a high level view of the network. This becomes extremely useful when then the user can also decide to increase or decrease the level of abstraction he wants to obtain. For example, the user can select a number of nodes using two modes of selection 1) lasso, and 2) click and drag to aggregate nodes of a small world network at multiple levels and see the details of a subgraph on a separate view.
Animation: it is a computerized information visualization technique that gives better clues about what the data relations are and help users relate two states of the system because it implicitly employs time as an extra dimension. For instance, when the user wants to animate the transformation of node-link diagram to adjacency matrix or vice-versa that involves some kind of interpolation of graphical elements from one state to another as shown in Fig. 11.
Figure. 11 The stages of an animation from a node-link diagram to an adjacency matrix
Matrix-based representation applications
It is a hybrid network visualization system that uses two representation 1) node-link diagram as shown in Fig. 12, and 2) matrix-based representation as shown in Fig. 13, it supports a lot of features for social network exploration such as interactive filtering and clustering, reorder (layout) matrices, different layout annotation and comparisons and supports consensus among several clustering. MatrixExplorer  attempt to fulfill social sciences researchers' needs for an exploratory system based on conducted sessions and questions.
Automatic ordering which tries to find an optimal linear order for all the vertices of the graph that propose a good initial matrix layout and to provide interactive tools to improve it, if needed. A "good" order, according to analyst, is one that reveals dense blocks and conversely avoids sparse isolated values.
Interactive layout that enables user to drag and move nodes, sort rows and columns according to one attribute, permute rows and columns circularly, allows locking a set of rows and columns together before reordering a matrix, filtering
Consensus among layouts where different layouts often imply different clusterings. It is important to be able to identify common clusters among layouts.
Figure. 13 MatrixExplorer Matrix-based representation view
Figure. 12 MatrixExplorer node-link diagram view
A potential cognitive cost of switching back and forth between the two representations
Techniques are poor in time complexity, because they need to compute the full distance matrix between all the vertices.
NodeTrix  integrates the best of the two traditional network representations by using node-link diagrams to visualize the overall structure of the network and matrix-based representation to show interconnections between nodes in communities. It solves a lot of existing problems related to graph drawing. For instance, in a community that is almost clique (a node is fully connected to all nodes in the network) and only missing few edges, one may suggest to just display missing edges, but this solution will not work out when the number of missing edges are equal to number of nodes. Matrix is a good choice for community analysis but it is lack of going back and forth between corresponding rows and columns, especially when it comes to networks that have tree structure and few communities where the determination to use either node-link diagrams or matrices is a problem, because choosing between these representations require a trade-off between readability of global structure and ease of community analysis.  represented node-link diagrams and matrices in two separated screens but with potential cognitive load and the distraction it cause from switching between representations. Other solutions have been proposed to improve the readability of node-link diagrams for communities such as aggregated node-link diagrams .
A readable representation for dense subgraphs where the underlying nodes and links serve as initial input for the network and aggregated networks are derived from them. Community contain either a group of the underlying nodes or a unique underlying node under one constrain underlying nodes are never shared by aggregate nodes.
Drawing links have three options: 1) displaying only aggregated links, 2) displaying only the underlying links and 3) displaying both where the benefit comes from the level of details the user needs to see between nodes, with the added flexibility of allowing the user to control the thickness of the links through a slider as shown in Fig. 13.
Moving a node or a matrix to adjust its position and improve the readability of the representation can be done by grabbing the matrix or the node, dragging it and releasing it at a new position. As the element is dragged, its connecting links are updated
Interactive selection: where the user may lassoselect the desired nodes, which are then immediately converted into a matrix, the user can add, extract, merge and split nodes or matrices.
On-demand axis labels, where the axis labels are not required on all matrices at all times and axis labels for individual underlying nodes may not be necessary at all in a final layout, so the user may hover with his mouse over the matrix and they will be displayed.
Exploration of Matrices: as we mentioned before, finding how two communities are connected is tedious as it requires going back and forth alternately reading rows and columns. Moreover, if communities are far apart in the matrix, this task requires a scan of the full length of matrix rows or columns, and connections in a large matrix may lie outside the viewport. Obviously, the task is worse when dealing with three matrices as the user needs to check for intersections of rows and columns in each of the three communities. With NodeTrix this problem is solved with a dual-view report.
Animation is supported in NodeTrix.
Doesn't support node duplicity where an actor is assumed just to belong to only one community, causing clustering ambiguity, whereas in real life an actor can belong to several communities. When an actor is connected to more than one community, this actor will be such a source for confusion to the network increasing the visual complexity of the graph by introducing node overlap and link crossings due to the tight space packing. Visualizing the overlapping nodes is difficult or even impossible when the number of intersections increases and which in turn will affect on the readability between communities. Nathalie et al  improved the readability of the clustered social network by using node duplication based on NodeTrix as shown in Fig. 14
Figure. 14 Clustered graph representation of the same portion of a co-authorship network. Orange marks represent authors, blue coauthorship and grey duplication links. (a) Clustered node-link diagram with communities circled. (b) NodeTrix representation with communities being adjacency matrices (authors are placed in rows and columns; blue squares indicate co-authorship). Missing intra-community links appear. (c) NodeTrix where actors shared among communities are duplicated in each. Missing inter-community links appear.
Figure. 13 NodeTrix drawing links
Most of the visualization tools we have mentioned are limited to uni-modal representations with various options (interactive views, calculating structural graph properties or deploying traditional statistical social network anaylsis) where each node represents a single data element type with an established relationship exist between nodes. With the changing structure and increased complexity of social network data, a more generic social network model is needed that supports multiple node, edge and descriptive types (mutli modal, relational and featured) as shown in Fig. 14. Lisa et al.  presented Invenio tool that bridges the gap between current social network visualization and graph mining software that address the complexity of the multi-modal social networks. Invenio support the creation of visual analytics views using database and graph mining operations.
Multi-modal interpretation of social networks has a number of advantages:-
Support a rich set of features for nodes and links in the form of pruning the dataset for visualization clarity, running an expensive graph mining algorithm for evaluation and prediction.
Easily to map between traditional graphs and relational model in database.
Allow the translation between different levels of abstractions.
Figure. 14 Multi-modal representation using Invenio
Data importing: Invenio supports the loading of data directly from a relational database.
User action history: recording the user actions can help streamline visual exploration
Graph mining that has structural properties degree and betweenness, two clustering algorithms, and a shortest path algorithm.
Network transformations that include different projections or graph representations of the multi-modal networks. For instance, the creation of a uni-modal network from multi-modal network.
The tool requires the support of more algorithms and structural properties.
The visual flexibility of the tool needs to be enhanced.
It doesn't work with networks that exceed 1000 node. More investigation need to be done with large graphs
There still one thing worth to be mentioned before moving to conclusion and future works, most of social networks contain millions of nodes and edges in real world examples and they tend to be scale-free networks as shown in Fig. 15, visualizing large graphs like these require some sort of simplification approach where a large portion of edges can be shown to be less important and removed while preserving node linking and other features such as cliques to be more clearly reveal the network's underlying connection pathways, the point of this simplification is to keep important node and edge semantics. Yuntao Jia et al  presented a novel deterministic filtering approach that improves the simplification, layout and visualization of scale-free networks. They found that graphs are rarely planar, and even the best layout methods yield a space-filling jumble of edge crossings for even medium-scale graphs. Clustering works well on planar graphs. But, when it is applied to non-planar graphs, it increase edge density which makes the layout less flexible and the display more jumbled with more edge crossings. So, the effect worsens with the increased clustering of larger graphs. Furthermore, the merged nodes and edges created by node clustering lose their original semantics .
Their method filters a graph by removing edges in order of increasing betweenness centrality. They constrain this filter to preserve connectivity and other features (e.g. cliques) by marking feature edges in a graph preprocessing pass, and keeping an edge if it is marked or if its removal disconnects a connected component of the original graph. The resulting simplified graph thus avoids the distraction of edges seldom utilities in the propagation of information across the network, while retaining the connectivity and other pre-identified features of the original. Furthermore, removing edges improves the flexibility, convergence and quality of node layout algorithms  as shown in Fig. 16.
Figure. 15 Unsimplified graph
Figure. 16 Simplified graphC:\Users\123\Desktop\23 simplification method.bmp
CONCLUSION AND FUTURE WORK
Graph visualization is a sub-field of information visualization. It focuses on visually representing abstract data elements and the relationships between and reduces the cognitive load to understand the global and local structure. In this survey, we presented two different graph layout representations (node-link diagrams and matrix-based representation), described what they are and highlighted their applications. We showed other visualization layouts with the need of simplification to improve visualization tasks. There still a lot of work to be done in the field of social network visualization, we offer here a personal in some open issues:-
Empirical studies for the effectiveness of graph layout algorithms: The problem of finding effective and efficient layouts is still open
Visualizations that match with substance
Adoption and further investigation of interaction techniques: deeper exploration of interactive techniques is needed
Coping with Internet-based communities and datasets: There is a need to scale up with large amounts of continuously updated information and current tools are still not ready to cope with such issues
Gaining confidence and coping with missing and incomplete data: Applications do not provide solid methods to gain confidence on the things one sees and to confirm or reject a hypothesis.