I'm using the well-known USC Burns Survival dataset to explore logistic regression in R.

The independent variable is burn area, and the outcome is binary survival (yes/no).

In the documentation, burn areas are grouped as the 'midpoint of set intervals', and taken as a log ie.

log(area + 1)

for sample datapoints like (highest/lowest):

1.35    yes
2.35    no

Medically, burns are usually specifed in terms of surface area % of total body eg. 40%, or alternatively, as an estimate of surface area in square meters. My question is: how does this relate to the dataset independent variable ie. what does '1.35' actually mean in terms of % body surface area burnt? What would eg. 30% burns become in the dataset, using the 'midpoint of set interval'?

Thanks guys

Ref: http://statmaster.sdu.dk/courses/st111/data/index.html#burns

  • $\begingroup$ When I clicked the link to the dataset, the access is forbidden. Is there any link to the burns dataset? $\endgroup$ – vtshen Jan 31 '18 at 2:51

OK, found the original article. 'area' in the model referred to third degree burn area in square centimeters. The independent variable 'midpoint' in the dataset is thus

X = ln(third degree burn area cm2 + 1)

Ref: http://www.stat.ncsu.edu/information/library/mimeo.archive/ISMS_1992_2089.pdf

  • $\begingroup$ Thanks for coming back to share the answer and don't hesitate to accept it (with the little check mark next to the vote number) to mark the question as “answered”. $\endgroup$ – Gala Jul 4 '13 at 14:51

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.