Dear friends,
We are looking for a set of sociology data suitable for doing nested logit regression model. However, each case must have records for all choice alternatives. For example, it is common in transportation studies that each individual have experienced all different transportation modes under study. We could not find any similar data set in the sociology discipline. Appreciate if anyone is willing share any suitable data set or articles' reference.
Many thanks.
Geoffrey.
Since you don't seem to care about what the variables are I'm assuming this is for student exercises. You might be able to use sub samples from the General Social Survey.
Dear Sam, many thanks to your suggestion. We shall look into it. Wishing you a happy New Year of 2017.
Dear Dr. Tso
I hope I got you right. You seem to intend to run multilevel logistic regression model using GLS. However, I am so sorry to inform you that you questionnaire design does not fall into the assumption for logistic regression, such as binary, ordinal and multinominal because your items are ipsative. You may either create dummy variables for the common answers if you would you like to conduct the multimonimal logistic regression. Nevertheless, there would trigger another problem: There would be too many different combinations and the pseudo R square may be so small. On the hand, you may try hierarchical partitioning analysis (HAP) which is often utilized in decision science.
I hope my words will help.
Happy New Year!
Dear Wing Yip, Thanks to your answer. If you are referring to the questionnaire data from General Household Survey, yes, we have to create dummy variables to run logistic regression. We are still trying. Thanks again and Happy New Year.
Dear Professor Sloss and Professor Chui,
With your experience, have you ever come across any application in sociology that use nested logit model? Many thanks.
Dear Dr. Tso
Thank you for your prompt reply. Indeed, I am just a Ph.D. student rather than a faculty member. I am very grateful for your appreciation.
On the other hand, with regard to your question, I personally believe that multinominal logistic regression may suit your situation. In addition, since your model is nested, I wonder if you are doing a multilevel study. If so, there will be many areas that your model can be applied to. For instance, I assume that General Household Survey, you may have two levels: 1. between-group level --> income groups 2. within group --> individuals. You may look into the interaction between WITHIN and ACROSS levels. You may also investigate fixed and random effects. There are different possibilities.
- Badges
- Science topic
- Similar topics
- Social Science
- Sociology