Subscribe to DSC Newsletter

Psychometric Methods and Applications


Psychometric Methods and Applications

Join our group to discuss psychometric methods such as latent class and profile analysis, latent regression, structural equation modeling, item response theory, cluster and factor analysis, scales development and validation, and more!

Members: 33
Latest Activity: Sep 21, 2014

This group was created with the purpose of facilitating social and professional communication between analysts in areas of media, marketing and political research, business intelligence, educational and professional testing, industrial-organizational psychology, and others. Content personalization, micro-targeting, measurement of engagement, dimensionality analysis, attitudinal and lifestyle measurement, segmentations and typologies, behavioral and attitudinal profiling, measurement and modeling of social networks, predictive modeling - these are only a few examples of problems and challenges that psychometricians and measurement practitioners across various areas and disciplines are facing in XXI century and that will (with your help!) be covered in our discussions. Members are welcome also to post announcements about upcoming workshops and conferences, discuss research ideas and projects, provide reviews for papers and software, and post links to interesting sites and sources!

Discussion Forum

Market Segmentation

Started by Konstantin Augemberg. Last reply by Tim Altier May 26, 2009. 1 Reply

Personalization of content and experiences

Started by Konstantin Augemberg Jul 18, 2008. 0 Replies

Comment Wall

Comment by Carol M Lazo on July 6, 2008 at 8:00pm
Hi there,

I must admit my background has been across the spectrum of qualitative and quantitative methods in application to understanding individual and group bx, though it is moving toward purely quant/inferential analysis.

In terms of typology development, what are the methodologies used in development, and applications? Can behavioral/profile segmentation be modeled at a regional level (i.e.: are there personality profiles that are inherent to specific cities/regions, and if so, how can they be quantified)?


Comment by Emory Creel on July 7, 2008 at 8:28am
Can I be extremely vague here? Two points: 1. similar tools (surveys, multinomial evaluation) and data types (multinomial, categorical, demographic) are used by both and therefore similar analytics are used. 2. In response to your question, there was a well-known book demonstrating the benefits of matrix charts
(a good example of this type of chart can be seen linked to the
showing judges' decisions. what interested me about this is that kennedy's responses were all in one direction, so we should be able to remove him (i.e. he offers no information), but then he was the deciding vote on 50% of the decisions. It shows how decision making methodologies can clash.). At any rate, this book had a matrix chart of policies of many countries, both european and american, and it interestingly demonstrated that policies in american countries were heavily linked (ie based-on)
policies of european countries with whom they were once closely associated. The king is dead, long live the king, I guess. But I think that answers your question from a very macroscopic level.

-emory creel
Hamilton Numbers
Comment by Konstantin Augemberg on July 15, 2008 at 10:33am

Great question! I would say that you may be looking at two different problems here. One is of a quantiative nature, known also as a a "measurement equivalence" problem. Just like with continuous scales (when you want to see your scale has the same dimensions and measure the same constructs across different samples and populations), in case of typologies (discrete latent structures), you want to see if your classification "holds" structurally across different groups (are there differences in "make-up" of segments across the groups).

Another way is to look at your typology from "qualitative" perspective. Now you are looking not at the structure of typology, but just how different are types across the regions of some country or other groups. So you are looking at the proportional differences across regions (e.g., using adjusted residuals) to see if in some regions, people tend to be of a particular "type".

Of course, typology does not have to be always generalizable. You can create typology/segmentation for a specificpopulation, e.g, typology of newspaper readers or typology of statistical analysts :)
Comment by Konstantin Augemberg on July 18, 2008 at 9:54am
What do you guys think about personalization and narrowcasting trends in media? I see three possible directions here:

- behavioral data-based personalization: system observes individual's behavior and adjust content, format and delivery methods based on those observations. For example, I go to news website; system analyses my clickstream and displays news articles and advertising in accordance with my past behavior. This method is often attacked by privacy advocates.

- "opt-in" personalization - when individual create a profile of content preferences and submits it to content provider. For example, I go to New York Times site and create "my NYT page" by selecting sections that I would like to see on my screen.

- profile-based personalization - you create profiles of people based on some third-party data (e.g, collected from personal interviews) and then re-create those profiles for people who use your system. For example, you interview people about their content preferences in newspapers, and then classify (segment) them based on their preferences. At the end, instead of one newspaper you now offer N different versions of newspaper.

Which of these methods do you prefer, and which one do you think will prevail in the future?


You need to be a member of Psychometric Methods and Applications to add comments!


Members (32)


On Data Science Central

© 2020 is a subsidiary and dedicated channel of Data Science Central LLC   Powered by

Badges  |  Report an Issue  |  Privacy Policy  |  Terms of Service