We used multiple regression to estimate the differences in total cost between the settings for birth and to adjust for potential confounders, including maternal age, parity, ethnicity, understanding of English, marital status, BMI, index of multiple deprivation score, parity, and gestational age at birth, which could each be associated with planned
place of birth and with adverse outcomes.12 For the generalised linear model on costs, we selected a γ distribution and identity link function in
preference to alternative distributional forms and link functions on the basis of its low Akaike's information
criterion (AIC) statistic.
Also encoded in the number are the person's dating
preferences, such as do they prefer to date a non-smoker, someone who likes pets, someone who wants children, someone who owns their own home, and so on, as well as the weight a person
places upon each of the different
criteria.
An automated computer system ideally «should» perform similar functions that a human appraiser would, but with quite a twist; it would not consider the human element regarding
preferences, being the very thing that humans consider when
placing personal value
criteria to a property