November 15, 2014

Learning Analysis of Social Networks

0 comments

This is my first look into social network analysis for learning.  We are starting with the idea that there is value in understanding how interactions happen during learning regardless of the context.  This sets the stage for digging in deeper and conducting analysis on the social networks that learners participate in such as twitter or a blog.

Dragan (our instructor) mentioned that researchers have often thought social networks may be the most important component of learning. And the analysis of social networks is based on various research fields. He mentioned some key characteristics that will be of focus include density, centrality, and modularity.

Network Elements
Social networks have some key structural elements that can be identified in order to establish a common language and conceptual model. This allows us to analyze them.  In this mooc we discussed three key elements, the actor, relations, and data sources.  

Actor
The actor is a node or vertex within the network. In social networks this is typically a person or learner, but I don’t think it would necessarily need to be a person.  

Relations
The relation in the network refers to the ties, edges, arcs, and links that connect the actors. Relations can be undirected and weighted or they can have a direction, meaning that an actor can be the sender and any actor that receives data can be the receiver. So, actors can be senders or receivers or both.

Additionally, the relation between two actors can also be labeled or categorized.  This means they can represent something, such as friendship, advice, hindrance, or can be a form of communication. I would imagine this could be a very interesting component of network analysis to try and identify and define these relations for the purpose of understanding the learner, the network, or the context.

Data Sources
The third element discussed in the mooc was the area of how the data was collected.  I think the idea of how you gather your data will have an impact on what filter you use to analyze the data.  Is it you own data such as email or is it from twitter?  This collection process impact what you can look at and create a potential bias on how you analyze the information and what conclusions you may be able to make from that analysis.  

As a reflection, I don’t think this third element is well articulated within the mooc materials.

Analysis
There were three key areas of analysis that we are looking at which include density, centrality, and modularity.   

Density is the degree to which actors are connected to all the other members in the network. Centrality is the extent in which the actors are organized around a central point.
Modularity is the way that you quantify the modules within the network or community, by counting and analyzing the ties between the actors. (It can get more complicated quickly, but this is the core.)

Potential Benefits
My first approach to considering how studying social networks might be beneficial to enhancing learning, is impacted by traditional classroom instruction. As a facilitator you often lurk about your classroom or learning environment and just listen.  What are they saying, who is saying it, who is dominating the conversations, and do they have a grasp of concept?  Should I step in and correct something that has been miss-communicated?  Do I need to provide clues to a small group to get them headed in the right direction?  Did they understand anything I said?

I would think if you could monitor a social network of a group of “actors” all with similar learning goals (i.e. a class or group of employees) you could begin to get a sense of how their learning was/is progressing. I would also think that it meet even create a safer environment for learners who struggle with social settings.  



November 1, 2014

Getting Started with Tableau

0 comments
Collecting Data

To begin working with Tableau, I decided to pull some West Virginia Department of Education data. The WVDE has a limited set of data that it makes available here.  The dataset that I grabbed was around graduation rates. I particular, looking at the percentage of students that graduated in 4 years.

The website defines it here....

defines a high school graduate as "a student who has received a regular diploma in either four years or five years as part of the four-year adjusted cohort or the five-year adjusted cohort." Students earning high school credentials by obtaining a high school equivalency diploma or a modified diploma are not considered graduates for the purpose of the graduation data.

Integration
There wasn't really any integration with the base data set.

Analysis in Tableau
The first step is to connect Tableau to your data. I did that be simple by connecting to the excel file. I did edit the column titles to help me once it was in Tableau.

(sorry iOS users this is in Flash, I will try to find a better tool)


Unable to display content. Adobe Flash is required.

In this video I walk through a very basic set of configurations to be able to see some data.  No real analysis, but it shows you what you can do pretty quickly.

The Learning Analytics Data Cycle Breakdown

0 comments
What is the Learning analytics data cycle?   

In our mooc week 2, @gsiemens presented the data cycle shown below. It represents a cycle of what you would need to do to work with and get insight from learning data, leading to understanding that should lead to an informed action.  Therefore, a data loop.  I will break down my understanding of the data cycle below.



Data Collection and Acquisition  
This is the specific process by which the data is collected. This is very critical first step. You can start by considering what data you have, but should very quickly start to consider what data you want, but don't have. If you don't figure out how to collect it, you won't get it.  

For example, do you want what time the learning activity started?  Not when it was scheduled, but when it actually started.  How do you get that information?  Does the data have to be auto-generated or is it acceptable to have a person manually take an action to collect the data?   For me, this is more trial and error then strategic thought, but I'd like to be more proactive in the future.  Can we create simple, reliable data collection points and gather all the data this already being generated?  

I am hopeful that I can start to map some of the Internet of Things (IoT) "offerings" to this stage. Can we put a sensor in the learning environment to help collect information?

Storage
Where will you store the data. This will quickly become another "big" consideration.  When you set out on an ideal journey, you want all the data to be stored in a single location.  But you quickly realize that this is just not possible. For many organizations, the LMS was put into place with this purpose, but all systems have limitations... i.e. when did the class really start?  9:02, 9:15... when did the students login?  Does that matter to you enough to store it in another system?   

Once your in a two system world, then you are starting to into the next phase....

Integration
Data integration can get tricky, you don't want to have to keep managing the your learner demographics, but depending on your organization, even who you are measuring can be pretty fluid. Learners can enroll, not participate, enroll late, leave the learner population all together, and a million of scenarios that can drive you crazy like sharing devices, leaving in the middle of the activity...  If you have a lot of data sources, this can quickly become challenging and complex.

This is were the Data Wrangling comes in. How do you start to get data that you can structure and use?  So far, for me, this has not been a fun activity.  It takes some patience. 

Analysis
Once you have integrated your data, you can begin to conduct your analysis of the what is happening in the data. In my world, you will typically have some ideas of what you want to seek out, at least initially, however once you start to (try) to answer questions, you can move into new areas, look for new trends, look for outliers, or correlations.

It is a little interesting, most best practices seem to point to some version of the scientific method, develop a hypothesis from which to work. What questions are you trying to answer?  Do students who show up 15 minutes late perform as well as those who were present at the beginning of the class?  As an educator, you hope not.  :) Just Kidding.  Of course, there are an endless number of questions and you should really spend some time writing them down, showing them to others, and try restating them differently. I also like to tell people to think about what action they are going to take when they get the answer? Kind of a "so what" test.

I am not yet that familiar with the concepts in the map above. SNA, NLP, Concept Development. I will have to do some research to see what these things can do for the analysis process.

Representation & Visualization
This process is about portraying your analysis in a why that enhances or at least explains your analysis. Can you see trending data? Can you see relationships in the data?  Can you compare volume or velocity?  Can you look at a map and quickly know what has happened or is happening?  

This can be a very powerful moment for a group of people working together. Again, spend time looking at the story with your team, peers, or force you significant other to look at it and tell you what they think. You will be surprised at what they tell you it means.

If you can start to see these things then it is time to take action.

Action
If you are engaged in Learning Analytics to "enhance learning" (see our original definition) then you are almost obligated to take action. What are the next steps?  Based on the data that you have, what actions can be taken to enhance learning?  How does the action get implemented?  Knowing what we know now (through the analysis) can we take action and continue to measure the impact of that action?

This will take us back to "data collecting"