Filtering Edges Relationships in Twitter

Jan 23, 2010 at 4:45 AM

Hello, hope you can help me - I haven't been able to find an answer for some reason, possibly just not looking in the right place.

I have extracted a small Twitter network, with the three Relationships of Mentions, Replies To, and Followed.

Two questions:

1. I assume each edge represented in the graph can be either one, or all, of the three relationships? How can you tell which one it is?

2. How can I filter them? For example, so my graph only shows Replies to, Mentions, or Followed? 
I tried filtering out Followed in the Dynamic Filter, but that seems to remove all the edges; and wouldn't solve the issue for Mentions and Followed
I tried Autofilter, and redoing 'Get Vertices from Edge list', but that doesn't seem to make a difference.
Copy-Pasting the relevant rows into another worksheet is an obvious solution, but it would be nice if there was a simpler way to filter them.

Thanks for your help,

Julian

Jan 23, 2010 at 5:34 AM

Julian:

1. No.  Each edge is one and only one of the three relationship types.  You can tell which one it is by looking at the Relationship column that gets added to the right edge of the Edges worksheet.  The values in that column are either Followed, Mentions, or Replies to.

2. You can filter the edges by clicking the down-arrow in the Relationship column header, then checking those relationship types that you would like to see.

You can also merge duplicate edges using NodeXL, Data, Prepare Data, Merge Duplicate Edges.  This will invalidate the Relationship column, but it will add an Edge Weight column that tells you how many relationships were merged into each merged edge.

-- Tony

Jan 24, 2010 at 4:17 AM

Thanks Tony,

I now see that what I was doing wrong was clicking on 'Lay Out Again' rather than 'Refresh Graph' when I used the autofilter. So I can now filter out the different relations.

However, not to be picky but just to understand properly - you say that "Each edge is one and only one of the three relationship types": I can see that easily in the worksheet, but on the graph it is different, if we have e.g.

A follows B
A mentions B

There is apparently only one edge on the graph, or are there in fact two but they overlay each other? If I apply the labels to the edges, 'Mentions' seems to be overlaid on other labels.

Also, while I'm at it. Now that I've applied the labels, I can't seem to get rid of them. I've tried resetting the Autofill Columns, and Refreshing the graph - what else should I do? I've had a similar problem with other settings (e.g. once you select a Scheme, there doesn't seem to be a 'Reset all and go back to the beginning option'; similarly, in the Autofill Columns, it would be good if there was an option in the drop down menu for 'Do nothing' - once you have selected something, the only option to empty that field is to 'Reset All'.

Thanks again for your time,

Julian

 

Jan 24, 2010 at 4:34 AM

Julian:

Sorry for the confusion.  Duplicate edges indeed overlay each other in the graph pane, so you cannot tell which is which.  We have talked about rerouting duplicate edges so they can be visually distinguished, but it turns out to be a difficult problem and we have postponed that work item.

One way to handle duplicate edges is to merge them (NodeXL, Data, Prepare Data, Merge Duplicate Edges), and then use Autofill Columns to fill the Edge Width column using the values in the Edge Weight column.  That causes the merged duplicate edges to be thicker than the non-duplicate edges.  Or you can merge duplicate edges and then autofill the Edge Label column using the values in the Edge Weight column.

There are a couple of ways to get rid of all edge labels.  The first is to clear the Label column in the Edges worksheet.  Because the Label column is hidden by default, you will have to show it first using NodeXL, Show/Hide, Workbook Columns, Labels.  The second way is to right-click the "Edge Label" text in the Autofill Columns dialog box, then select "Clear Edge Label Worksheet Column Now."  (Yes, I know this is hopelessly hidden.  We will fix that in a future release.)

There is currently no way to clear a scheme except to manually clear the columns filled in by the scheme.  To clear the Weighted Graph Scheme, for example, you would have to manually clear the Color, Width, and Opacity columns in the Edges worksheet.

-- Tony

Jan 24, 2010 at 4:47 AM

Excellent! Problem solved :)

Thanks for the rapid response! Much appreciated.

Cheers,

Julian