Usage Data
Locked
 
|
Hello, I am trying to figure out the easiest way to handle this data:
I have 6 variables that represent different aspects of website usage for 32,000 people.
For example, one variable is health. Another is lifestyle. Another is family. Usage on a variable can range from 0 (no clicks on that category) to thousands of clicks on a category.
However, what I'd like to show is the different variable combination groups...totaling to the N of the sample.
For example, how many people launched health only? Lifestyle only? Family only? Health and lifestyle only? Health and family only? Health, lifestyle and family only?.....and so forth and so on....until I have isolated all 32,000 users within one of the
groups.
What is the best way too do this analysis?
Thanks in advance!
Tiffany
Tiffany Perkins-Munn, Ph.D.
Adjunct Professor of Psychology William Paterson University Wayne, New Jersey [hidden email] ===================== To manage your subscription to SPSSX-L, send a message to [hidden email] (not to SPSSX-L), with no body text except the command. To leave the list, send the command SIGNOFF SPSSX-L For a list of commands to manage subscriptions, send the command INFO REFCARD |
|
|
The fancy way is to use log-linear modeling for 6 variables. Otherwise:
=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD
With only 6 variables, there are only 64 ( = 2^6 ) exclusive categories, so you can create a variable with 64 categories and tabulate them. I would probably look at the most frequent 3 or 4 variables in order to focus on 8 or 16 categories. Then I would look closely at whether there is anything interesting happening with the ones left out. COMPUTE top3= 100*var1 + 10*var3 + var6 . -- Rich Ulrich Date: Wed, 16 Jul 2014 18:48:48 +0000 From: [hidden email] Subject: Usage Data To: [hidden email] Hello, I am trying to figure out the easiest way to handle this data:
I have 6 variables that represent different aspects of website usage for 32,000 people.
For example, one variable is health. Another is lifestyle. Another is family. Usage on a variable can range from 0 (no clicks on that category) to thousands of clicks on a category.
However, what I'd like to show is the different variable combination groups...totaling to the N of the sample.
For example, how many people launched health only? Lifestyle only? Family only? Health and lifestyle only? Health and family only? Health, lifestyle and family only?.....and so forth and so on....until I have isolated all 32,000 users within one of the
groups.
What is the best way too do this analysis?
Thanks in advance!
Tiffany
Tiffany Perkins-Munn, Ph.D.
Adjunct Professor of Psychology William Paterson University Wayne, New Jersey [hidden email] ===================== To manage your subscription to SPSSX-L, send a message to [hidden email] (not to SPSSX-L), with no body text except the command. To leave the list, send the command SIGNOFF SPSSX-L For a list of commands to manage subscriptions, send the command INFO REFCARD |
|
|
how about a cluster analysis?
On 7/16/2014 7:28 PM, Rich Ulrich
wrote:
-- Stephen C. Peck Assistant Research Scientist Achievement Research Lab Research Center for Group Dynamics Institute for Social Research University of Michigan 426 Thompson Street, # 5136 Ann Arbor, MI 48109-1290 (734) 647-3683; fax (734) 936-7370 http://www.rcgd.isr.umich.edu/garp/ [hidden email]===================== To manage your subscription to SPSSX-L, send a message to [hidden email] (not to SPSSX-L), with no body text except the command. To leave the list, send the command SIGNOFF SPSSX-L For a list of commands to manage subscriptions, send the command INFO REFCARD |
|
In reply to this post by Perkins, Tiffany
I think there's two steps involved here:
1) dichotomize the numbers of clicks. You proposed 0 versus 1(+) clicks per category but you could also consider median splits which can be easily created with RANK. For some examples, see http://www.spss-tutorials.com/rank/. 2) if you apply suitable value labels to these dichotomous variabels, you can combine them by concatenating their VALUELABELS into a (new) long string variable. Optionally, AUTORECODE that. For an example and some backgrounds on this approach, see http://www.spss-tutorials.com/combine-dichotomous-variables/. HTH, Ruben |
|
In reply to this post by Perkins, Tiffany
I am not sure exactly what you are looking but for one way of reading your post, you could coarsen your measures and use MULT RESPONSE or CTABLES.
One way to coarsen is to do something like RECODE health lifestyle family (0=0) (1 thru hi=1)(else=copy) into health2 lifestyle2 family2. Other recodes would give you different ways to coarsen your measures. e.g., RECODE health lifestyle family (0=0) (1 thru 10=1)(11 thru hi=2)(else=copy) into health2 lifestyle2 family2.
Art Kendall
Social Research Consultants |
|
|
In reply to this post by Perkins, Tiffany
I am not sure exactly what you are looking but for one way of reading your post, you could coarsen your measures and use MULT RESPONSE or CTABLES.
One way to coarsen is to do something like RECODE health lifestyle family (0=0) (1 thru hi=1)(else=copy) into health2 lifestyle2 family2. Other recodes would give you different ways to coarsen your measures. e.g., RECODE health lifestyle family (0=0) (1 thru 10=1)(11 thru hi=2)(else=copy) into health2 lifestyle2 family2.
Art Kendall
Social Research Consultants |
«
Return to SPSSX Discussion
|
1 view|%1 views
| Free forum by Nabble | Edit this page |