All,
I have a data-set which contains information on both the type of property an individual(s) purchased, along with some information on the individuals themselves.
However, I am missing age related information for a number of years and I need to account for this as I think not thinking it will bias results. My data is not missing at random - I am aware that the missing cohort are generally older.
The variables I'm using in my model to predict age include; region the property was purchased, price of the house, whether the person was a first time buyer, whether the property was a house or apartment. When I run the model I'm getting an R2 of 0.26.
I also compare the median age of records that I actually have a value for, along with the predicted age. I have a median age of 39 for the actual age as opposed to 46 for the predicted age.
Is there any literature out there that has looked at attempting to predict the age of home buyers?
Hi Kelevyn,
It's basically a research paper on the characteristics of home owners. So one chapter will focus on the age of buyers, nationally and regionally. With missing values at present I have a downward bias in the median age of buyers for early years. So without accounting for this I might say "The median age of buyers increased by 8 years over the period". However, I know I'm missing the ages of a generally older cohort so in actuality it may have only increased by 6 years.
OK, understood - you may want to look at http://www.missingdata.org.uk/ ; it has a discussion forum.
The other thing you could do is to do a weighted analysis where you inflate the age profile to the national or overall picture ; most statistics software allows weights. https://surveyinsights.org/?p=6255
Lavallée, P. & Beaumont, J.-F. (2015), Why We Should Put Some Weight on Weights. Survey Insights: Methods from the Field, Weighting: Practical Issues and ‘How to’ Approach, Invited article, Retrieved from https://surveyinsights.org/?p=6255
- Badges
- Science topic