Christina's LIS Rant
Sunday, November 12, 2006
Unexpected overlaps: Qual research and SNA
I'm doing things backwards in my doctoral program. You're *supposed* to start with quantitative methodology classes and then take the other things later... Here I am 2/3 through my coursework and I won't be taking my first quantitative methodology class until next Fall. Anyway, both of my classes this semester have spent a good deal of time differentiating themselves from quantitative methodology. SNA (social network analysis) by virtue of the fact that the analysis is on relationships, not on attributes. Qualitative methods -- well they're just completely different :)
So, imagine my surprise when doing the reading this week when the Qual book cites Borgatti and they both discuss the same methods!
In my intro to qualitative research class, I'm reading chp 7 of the book linked above, which is "Data Management and Analysis Methods" by Ryan and Bernard. In my SNA class, I just finished reading chp 8 of Wasserman and Faust which is on affiliations and subgroup overlaps.
The standard SNA methods work on square matrices in which the rows and columns are the same actors. Affiliation or membership matrices are not square and have actors as the rows and events/clubs, etc., as the columns. Most easily, you can decompose the affiliation matrix into square matrices and look only at co-memberships between the actors (X^N=AA') or overlap of events (X^M=A'A). You lose a lot of information this way, but we know well how to deal with matrices like this.
So, in the qual book, I read about structural analysis and semantic networks looking at study by Nolan and Ryan in which they looked at terms participants used to describe horror films and then decomposed this to a person by person similarity matrix to look at co-occurrence of terms. I guess this is really actually the same thing as the SNA, it's just that they're talking about how the data is gathered and coded -- we start with data in hand in the other class.
Both books also talk about multidimensional scaling and correspondence analysis, too, but I think I need a bit more help to see how this makes things clearer -- to me it looks more muddy in both worlds.
Anyway, just a random thought on a rainy, cold Sunday. ¶ 1:08 PM| |cites (technorati) |
So, imagine my surprise when doing the reading this week when the Qual book cites Borgatti and they both discuss the same methods!
In my intro to qualitative research class, I'm reading chp 7 of the book linked above, which is "Data Management and Analysis Methods" by Ryan and Bernard. In my SNA class, I just finished reading chp 8 of Wasserman and Faust which is on affiliations and subgroup overlaps.
The standard SNA methods work on square matrices in which the rows and columns are the same actors. Affiliation or membership matrices are not square and have actors as the rows and events/clubs, etc., as the columns. Most easily, you can decompose the affiliation matrix into square matrices and look only at co-memberships between the actors (X^N=AA') or overlap of events (X^M=A'A). You lose a lot of information this way, but we know well how to deal with matrices like this.
So, in the qual book, I read about structural analysis and semantic networks looking at study by Nolan and Ryan in which they looked at terms participants used to describe horror films and then decomposed this to a person by person similarity matrix to look at co-occurrence of terms. I guess this is really actually the same thing as the SNA, it's just that they're talking about how the data is gathered and coded -- we start with data in hand in the other class.
Both books also talk about multidimensional scaling and correspondence analysis, too, but I think I need a bit more help to see how this makes things clearer -- to me it looks more muddy in both worlds.
Anyway, just a random thought on a rainy, cold Sunday. ¶ 1:08 PM| |cites (technorati) |
Comments:
Post a Comment
