Abstract
This course will concentrate on statistical methods to analyze local and distant failure after radiation therapy, with examples drawn from breast cancer (and candy packaging and fairy tales). When all patients have been followed T years, the T year failure percents in different sites follow a multinomial distribution, leading to Fisher's exact test for treatment differences and polycotomous logistic regression for prognostic factors. When failure rates are very low and the interest is in average rate per patient year, Poisson tests and regression can be used (named for Simeon Denis, not Roger). When time to first failure is of interest, at least some patients still haven't failed and the time of observation on these patients is statistically independent of their ultimate time to failure, actuarial methods can be used (e.g., logrank and modified Wilcoxon tests, proportional hazards and Weibull and exponential models). The assumptions required by each of these tests and models will be reviewed, as well as ways to decide if a particular study reported in the literature violates the assumptions.
Citation Formats
Gelman, Rebecca.
Biostatics.
United States: N. p.,
1995.
Web.
doi:10.1016/0360-3016(95)97621-7.
Gelman, Rebecca.
Biostatics.
United States.
https://doi.org/10.1016/0360-3016(95)97621-7
Gelman, Rebecca.
1995.
"Biostatics."
United States.
https://doi.org/10.1016/0360-3016(95)97621-7.
@misc{etde_20391306,
title = {Biostatics}
author = {Gelman, Rebecca}
abstractNote = {This course will concentrate on statistical methods to analyze local and distant failure after radiation therapy, with examples drawn from breast cancer (and candy packaging and fairy tales). When all patients have been followed T years, the T year failure percents in different sites follow a multinomial distribution, leading to Fisher's exact test for treatment differences and polycotomous logistic regression for prognostic factors. When failure rates are very low and the interest is in average rate per patient year, Poisson tests and regression can be used (named for Simeon Denis, not Roger). When time to first failure is of interest, at least some patients still haven't failed and the time of observation on these patients is statistically independent of their ultimate time to failure, actuarial methods can be used (e.g., logrank and modified Wilcoxon tests, proportional hazards and Weibull and exponential models). The assumptions required by each of these tests and models will be reviewed, as well as ways to decide if a particular study reported in the literature violates the assumptions.}
doi = {10.1016/0360-3016(95)97621-7}
journal = []
issue = {971}
volume = {32}
journal type = {AC}
place = {United States}
year = {1995}
month = {Jul}
}
title = {Biostatics}
author = {Gelman, Rebecca}
abstractNote = {This course will concentrate on statistical methods to analyze local and distant failure after radiation therapy, with examples drawn from breast cancer (and candy packaging and fairy tales). When all patients have been followed T years, the T year failure percents in different sites follow a multinomial distribution, leading to Fisher's exact test for treatment differences and polycotomous logistic regression for prognostic factors. When failure rates are very low and the interest is in average rate per patient year, Poisson tests and regression can be used (named for Simeon Denis, not Roger). When time to first failure is of interest, at least some patients still haven't failed and the time of observation on these patients is statistically independent of their ultimate time to failure, actuarial methods can be used (e.g., logrank and modified Wilcoxon tests, proportional hazards and Weibull and exponential models). The assumptions required by each of these tests and models will be reviewed, as well as ways to decide if a particular study reported in the literature violates the assumptions.}
doi = {10.1016/0360-3016(95)97621-7}
journal = []
issue = {971}
volume = {32}
journal type = {AC}
place = {United States}
year = {1995}
month = {Jul}
}