Before You Start:

Because you will not be collecting data in this lab, feel free to knit early and often to see how your responses are being formatted! Please do your best to maintain the formatting provided by this assignment. It makes grading significantly easier when answers are easy to read.

Download all the files for this lab and save them in the same folder. Open the CSS_Lab.R file in RStudio (File > Open). After the R script is loaded in the editor, set the working directory so that R knows where to find the RData file you are going to load (Session > Set Working Directory > To Source File Location).

The data file krackhardt_css_data.RData consists of two CSS data objects:

● advice_nets: respondents’ perceptions about their own and others’ advice ties within the organization

● friendship_nets: respondents’ perceptions about their own and others’ friendship ties within the organization

Analysis

Conceptually, how do these two networks differ from one another? What are the pros and cons of using this method? The network are very different by LAS from column method and row method. For LAS from column method, the network is sparse, there is really no node has high centrality of degree. While for LAS from row method, node 11 and has high outdegree, he/she percerives to be friends with other people, while other people doesn’t think so. Pros: Row method represent one’s own idea with network, it represents one’s true feeling about network. This method is subjective and could be potential more accurate if people’s feeling about their network are correct. Cons: However, people might not answer survey question honestly, this could cause inaccurancy for network data.

Analysis

What information does the ‘intersection’ method capture? What are the pros and cons of using this method? The edge only exists if two people both agree they are connected. This represents the will of people’s feeling about connection. Pros: the imformation about if two people connect(like “1” in this matrix) is highly correct. Cons: If one people mentioned the friend relationship with another people but another people didn’t mention about it, then this edge will not be counted. This cause problem. This concept is too strict. Sometimes people may feel difficult of determining who are truly friends or just working partners. Also some person didn’t confirm the friend relationship is because they are too cool to admit it. Thus intersection method could potentially reduce number of edges.

Anaylsis

What kind of information does the union method capture? What are the pros and cons of using this method? The union method will show that number of friends in network would be more than true story. As if one people believe there is friendship between two people, then they are friends(“1”) in matrix no matter how other people disagree.

Pros: It should capture non-friend relationship correctly in this case. Cons: It captures too many frind relationships which cause chaos for analysis.

Analysis

What kind of information does the median measure capture? What are the pros and cons of using this method? Median says if more than 50% of people believe two people are friends, then they are friends. Pros: It combines most people’s observation and feel, its dataset is potentially reliable. Cons: Sometimes, what most people believe could be wrong especially when they are not familiar with those people.

Finally, we’ll also load the data for our advice network for later analysis.

ad_column <- consensus(advice_nets, mode="digraph", diag=FALSE, method="OR.col")
ad_row    <- consensus(advice_nets, mode="digraph", diag=FALSE, method="OR.row")
ad_intersection <- consensus(advice_nets, mode="digraph", diag=FALSE, method="LAS.intersection")
ad_union  <- consensus(advice_nets, mode="digraph", diag=FALSE, method="LAS.union")
ad_median <- consensus(advice_nets, mode="digraph", diag=FALSE, method="central.graph")

Analysis

Describe what this network shows in your own words. For nodes, this network shows that nodes 1, 2, 21, 19 and 14 have high indegree connections, many people are believed to be friends with them. node 11 has high outdegree which means node 11 believe he/she is friend of many people. Node 18 doesn’t have friends. Also there are many mutual dyadic edges, and also triads such as 5, 14, 19 and 14, 19, 3. It follows theory of balance.

Friendship, Row.

Analysis

Describe what this network shows in your own words. In this network, still node 11 has most outdegree connections. Node 2 and 14 seems to have highest indegree connections. As it is row method, the arrow direction of edge are all personal feelings with other people, which results into many asymmetric edge. Node 18 and 20 have no friends. Node 10 can also be preceived to have no friends as no arrow is pointed to him.

What are the relevant similiarities and differences between the two networks? What do they mean? The first network is union method and second network is row method. similarities: Both networks have some nodes with high degree centrality such as nodes 2 and 14. Also, nodes 10 and 18 both have no friends. differences: There are more links in first network. Also first network has much more mutual edges.

As the union method includes all edges of row method.They potentially have similar layouts.

Analysis

Describe what this network shows in your own words. This network uses the intersection method. By intersection method, it is found that half of nodes are not connected with other nodes. Even some nodes are connected, the network is very sparse. But the connected edges shall have high accuracy. Node 1 has highest centrality in this network. Among 9 edges, 5 of them are mutual edges and 4 of them are asymmetric edges. There is no triad case in this network.

Advice, Median.

Analysis

Describe what this network shows in your own words. This network is generated by median network. This is the network that more than 50% of people perceive(agree on). In this network, half of nodes have no connections. The network is very sparse and there are no triads.

What are the relevant similiarities and differences between the two networks? What do they mean? similarities: both networks are very sparse, similar number of edges. differences: Two methods have very different edge connections. e.g. By intersection method, node 1 has highest degree of connection, while by median method, no one is connected to node 1.

Analysis

What respondents did you choose to visualize? Why? I choose respondent 18 for both advice seek and friendship visulization. As node 18 is disconnected with other nodes in networks. I’m intereted how node 18 would draw the network and try to find out why he/she is isolated.

What do their networks show? Can you draw any conclusions about each actor’s role in the network? For friendship network, the edges are very sparse. We could guess that node 18 doesn’t have any friends, also he doesn’t know many people and their relations, thus only 5 edges present. For advice network, it is a very dense network, most of nodes receives a large amount of indgree edge except node 9, 15 and 16. The advice network is pretty active. Thus some people in this company are mainly in work relationship. Also most of edges in advice seeking network are mutual network, people looks for help in mutual way. It is difficult to draw conclusions about actor’s role in these two networks, but I could guess node 9, 15 and 16 are not the core person in this company as they didn’t play important role in advice seeking network.Also node 18 might be manager of this company, though he/she doesn’t have any friend, still many people come and ask him/her for advice.

Analysis

What does this network show? Why might this visualization be useful? This network shows that node 1, 21 and 14 has highest degree of connections. Node 9, 10, and 17 has only a few connections. Node 18 and 13 has no connections. The edge in this visulization represents people who are friends and also advice seeking. This network can be compared to friend network aggragated by mutual method, it is found that the edge number of node 19, 13 drop significantly, which means they don’t ask advice from all of their friends.

Analysis

Outline your rationale for choosing your networks. I’ll choose structure of row aggregation. The reasons I choose row method is that: It has relative correct information about edge, and it doesn’t have many edges like union do but could potentially show correct results. Also, unlike intersection which removes a lot of edge information, the remaining edge is only a small set.

Analysis

How do you interpret high and low values in this matrix, calculated using Euclidean distance? The high value means that two nodes has significantly different structure equivalence. While low value indicates that two nodes have similar structure equivalence.

Which network has the smallest minimum distance between nodes? Why might that be? You may want to refer to your earlier visualizations for more insight into the network. The advice network, the nodes are very densely connected. It shortens the travel distance between nodes.

Which network has the greatest maximum distance between nodes? Why might that be? The friend network, the network are more sparse and some of the nodes are not connected. The maximum distance between unconnected nodes are infinite large.

Which network exhibits more structural equivalence? Friend network.

Analysis

Examining the results from above, which respondent’s perceptions were the most/least “accurate” when compared to the median response (assuming the consensus is the ground truth)?66 Hint: look for the strongest correlation between the respondent’s network and the median network. According to results above, respondent 16(0.7) is most accurate and respondent 1 is least accurate(0.28).

Analysis

Are the results for the friendship network very similar or different from those you saw in the advice networks? Give some possible reasons why individuals have more precise representations of one kind of relation structure than another kind of relation structure. The results are very different. In advice network, the highest/lowest coefficients of correlation are both higher than friend network, this indicates better prediction of advice network. Also, the highest correlation coefficient in advice network is however lowest in friend network. People have more precise prediction of advice network, the reason could be that people could see when one persons ask advice or chat with other people. But they are not necessary friends. This surface level observation makes people confused about relationship between them. Thus friend network are more difficult to predict.

Analysis

Which centrality score of individuals in the consensus network are most highly correlated with their accuracy in predicting the consensus network? Based on the readings, suggest a rationale why individuals’ embeddedness or patterning of ties might result in different perceptions. The degree centrality of individual are most highly correlated with their accuracy in predicting consensus network. The ties of individual will influence that persons’s perceptions about network. Like what his/her friends think or who his/her friends make friends will all have influence on his/her thoughts.

Analysis

Which individual sees the two networks as the most similar? Which sees them as the least similar? respondent 16 sees two networks as the most similar, as it has highest coefficient of correlation: 0.47 respondent 18 sees two networks as the least similar, as it has lowest coefficient of correlation: 0.03