Search the Community
Showing results for tags 'cluster analysis'.
Found 4 results
Hey guys! I'm relatively new to NVIVO and I'm trying to do a cluster analysis to show interconnected some nodes are. I have 53 nodes based on expert interview data (characterizing barriers to electric vehicles), and I'm trying to show that there's four main barriers that relatively linked to each other. I created a cluster analysis, see attached. The problem is that all the nodes seem to be of the same importance because they're the same size. However some nodes have 100's of source material, where as others have only one. Does anyone know if you can change the size of the nodes in the cluster analysis by the amount of times something is coded within that node? Secondly, I know that many of the sources are coded in more than one nodes (for example "Range" and "Price" were often coded together). They're close to each other, but they don't show how many times they are coded similarly. Is there a way to add a line between nodes that have certain level of overlap? Thanks in advance for your help!
The two options for cluster analysis appear to be Source and Node. Is it possible to do a cluster analysis for sources that I've grouped using case classification attributes? For example, let's say School 1, School 2, School 3, ... n each have a # of sources (records, interviews, etc.). I made these sources into cases, then assigned the names of schools as case classification attributes (so School 1 is the attribute value for all its sources and so forth). I want to see which schools (= its collection of sources) have similar coding.
Hello, I ran a cluster analysis on sources of 370 text files, which were clustered by word similarity. The results generated were 65,536 pairs (i.e., 65,536 Pearson correlation coefficient). However, there should have 68,265 comparison pairs (369*370/2). About 2,729, or 4%, pairs were missing, and it would take a long time trying to figure out which pairs were missing. Your suggestions to fix this problem will be greatly appreciated. Thank you, YT Chang
I am writing my doctoral thesis and I´m worried about some methodological issues dealing with the way Nvivo handles cluster analysis. Exactly I wonder how it handles data. I don´t feel prepared to answer questions about this issue to my examination comitee if they would want to know. I don´t mean the basis of cluster analysis that can be found in a statistics manual, but the choices the software make after the user click the "finish" button. I would like to know the technical details about how Nvivo handles cluster analysis in order to be able to interpret results correctly as well as properly address ussues of validity. I´m talking about something like a white paper. Not the kind of explantion that can be found in the software help files or tutorial, but the kind of information about operations the programer used to write the code and algorytms that process the data. I mean what "choices" does the software do after I chose to work with resources/nodes and analyse words/coding similarity?. I would like to read an official QSR document about this issue that I can cite and refer to in case I would be asked, rather than a short informal post. For example, when the user chooses "word similarity" what words are considered? Every word o are some words ignored, like "blind words"?. And when the user chooses "coding similarity" what nodes are considered? All, including cases nodes, matrix nodes, relationships, query results and everything?. If the user wants to restrict the comparison to just some nodes, can it be done?. I wanted to do it so I created some temporal case nodes containing only the passages coded at the nodes I wanted to restrict the comparison to. Then I used these nodes to do the cluster analysis. What nodes are compared in the analysis in this case?. Also I would like to read what other users think of and/or address issues of vality of cluster analysis of qualitative data with Nvivo.