Thanks for sharing Bruce. Can I ask what statistical methods were used to compensate for the inherent biases in social circle surveying? I couldn’t see it in the methodology document.
'I'm not sure which "inherent biases" you refer to in social circle surveying, but the methodology creates what's called an implicit super sample, which means people on average think of 20 people when they generalise a prediction, and demographic distribution is instinctively inherent as they aggregate the views of everyone they know, which includes many people who don't take online surveys. When you in turn aggregate all those predictions, you get a very good representation of a general estimate.'
Thanks for replying and sharing the reference. I hadn’t read that paper. My question was from a place of interest and the inherent biases are the ones of egocentricity and homophily. I have not used this method before so was interested if any calculations are used to correct these biases to improve validity of results. I heard a lot in the recent US election coverage of statistical models to compensate for the consistent under sensitivity to trump votes. I wondered if this study used any of those? Or if those are used exclusively for self report studies.
Thanks for sharing Bruce. Can I ask what statistical methods were used to compensate for the inherent biases in social circle surveying? I couldn’t see it in the methodology document.
Alfred Malmros from Early Studies replies:
'I'm not sure which "inherent biases" you refer to in social circle surveying, but the methodology creates what's called an implicit super sample, which means people on average think of 20 people when they generalise a prediction, and demographic distribution is instinctively inherent as they aggregate the views of everyone they know, which includes many people who don't take online surveys. When you in turn aggregate all those predictions, you get a very good representation of a general estimate.'
In science language:
https://www.sciencedirect.com/science/article/abs/pii/S0261379421001451
Thanks for replying and sharing the reference. I hadn’t read that paper. My question was from a place of interest and the inherent biases are the ones of egocentricity and homophily. I have not used this method before so was interested if any calculations are used to correct these biases to improve validity of results. I heard a lot in the recent US election coverage of statistical models to compensate for the consistent under sensitivity to trump votes. I wondered if this study used any of those? Or if those are used exclusively for self report studies.