What are the patterns of image diffusion during the Nepal Earthquake?
by Karen Blakemore, Brian McKean and Mingqi Liew

Questions

What is the relationship between number of followers and number of retweets? (contributed by Brian McKean)

What is the local and international percentage of retweet and what is the proportion of that? (contributed by Mingqi Liew)

Do images from influential tweeters live longer? (contributed by Karen Blakemore)

The chart below shows how the lifespan of an image varies based on the number of followers of the original tweeter, with an origin at the upper left corner.
For all images where the originating tweet is also in the data set, you can see a point on the graph. There are 186 points on the graph.
The x-axis is the tweet lifespan and ranges from 0 to 127,816,000 seconds which is 148 days. Where there is more than one tweet originating an image, the lifespan is measured from the oldest tweet. In eighteen cases, there was an originating tweet for an image, but its' date was greater than that of the retweet. 105 of the datapoints have an x-coordinate of zero, indicating that the corresponding image was not retweeted.
The y-axis is the sum of all the followers of the original tweeters. It ranges from 1 to 74,383.
# Findings, Recommendations and Future Work ## Local vs. International Reweets The percentage of local tweets which are retweets is 86.5%, whereas the international percentage is 97.7%. This indicates that tweets sent from Nepal are more likely to be original tweets than tweets sent from outside of Nepal. ## Association between Follower Count and Diffusion The analysis shows a relationship between the orginator follower count and the number of retweets. The association with image lifespan is less conclusive. We wonder how representative the data set is of the image tweet stream. A large proportion of the images (243/429) did not have the originating tweet in the data set. Of those images which did have the originating tweet in the data set, a large proportion (105/186) were not retweeted. It is our understanding that the tweets were collected based on keywords and hashtags. If the originating tweet was sent before the collection started, it would not be included in the data set. Because of this, we recommend attempting to gathering additional tweets of the images from the Twitter database, using the image id, if this is possible. As follow-up work, we suggest a similar analysis of the retweeters' follower count and its influence on image diffusion. We also recommend viewing a sample of images from the data set to determine if there is commonality between those that were retweeted and those which were not.