***** To join INSNA, visit http://www.insna.org *****
I collected a matrix of 86 fashion stores in Berlin.
Each has a number of attributes, of which I extract
only three (quarter, style and price)
First, I am able to determine possible correlations
between the attributes (quarter/style, quarter/price,
style/price). As a matter of fact, these correlations
are hard to find in my data (and in the city itself as
well, I find).
My goal is to determine what I would call "similarity
hierarchies", showing which stores are alike (or
"Mainstream, if you will) regarding the three
qualities and which are rather "different" (an
adjective used quite often in Berlin, especially
regarding the "vane" field from which the data stem).
Of course the data structure is simple, but maybe
useful to make some points (also to those who are
relatively new in the field like myself).
What I do is this:
I treat the matrix as 2-Mode data (86 stores as
actors; 10 quarters, 6 styles, 6 price range as
events). For instance:
Store A, Berlin-Kreuzberg, Casual, affordable
Store B, Berlin-Charlottenburg, Classical, high
Deriving a 1-mode network allows me find correlational
data between the 86 stores.
I apply the valued core function, in this way taking
into account not only the number of arcs but also the
Now there´s the bias question. If I picked only 2
stores in Charlottenburg (out of available 15) with
affordable prices, Store B holds a lower valued core
measures and might be labelled "less mainstream" than
In order to normalize the data, I calculate a factor:
The inversed sum of (Stores in the quarter + Stores
with this particular style + Stores in this particular
Multiplying the valued core measure by the
normalization factor I obtain a measure of the
"similarity" or "uniqueness" of the fashion stores.
Any thoughts on this approach are appreciated. I am
willing to provide a summary of the answers to the
list, if you prefer to answer to me directly.
Thanks in advance,
Do You Yahoo!?
Tired of spam? Yahoo! Mail has the best spam protection around
SOCNET is a service of INSNA, the professional association for social
network researchers (http://www.insna.org). To unsubscribe, send
an email message to [log in to unmask] containing the line
UNSUBSCRIBE SOCNET in the body of the message.