A novel high-power test for continuous outcomes truncated by death
Research output: Contribution to journal › Journal article › Research
Standard
A novel high-power test for continuous outcomes truncated by death. / Jensen, Andreas Kryger; Lange, Theis.
In: arXiv.org: Statistics, 27.10.2019.Research output: Contribution to journal › Journal article › Research
Harvard
APA
Vancouver
Author
Bibtex
}
RIS
TY - JOUR
T1 - A novel high-power test for continuous outcomes truncated by death
AU - Jensen, Andreas Kryger
AU - Lange, Theis
PY - 2019/10/27
Y1 - 2019/10/27
N2 - Patient reported outcomes including quality of life (QoL) assessments are increasingly being included as either primary or secondary outcomes in randomized controlled trials. While making the outcomes more relevant for patients it entails a challenge in cases where death or a similar event makes the outcome of interest undefined. A pragmatic - and much used - solution is to assign diseased patient with the lowest possible QoL score. This makes medical sense, but creates a statistical problem since traditional tests such as t-tests or Wilcox tests potentially looses large amounts of statistical power. In this paper we propose a novel test that can keep the medical relevant composite outcome, but preserve full statistical power. The test is also applicable in other situations where a specific value (say 0 days alive outside hospitals) encodes a special meaning. The test is implemented in an R package which is available for download.
AB - Patient reported outcomes including quality of life (QoL) assessments are increasingly being included as either primary or secondary outcomes in randomized controlled trials. While making the outcomes more relevant for patients it entails a challenge in cases where death or a similar event makes the outcome of interest undefined. A pragmatic - and much used - solution is to assign diseased patient with the lowest possible QoL score. This makes medical sense, but creates a statistical problem since traditional tests such as t-tests or Wilcox tests potentially looses large amounts of statistical power. In this paper we propose a novel test that can keep the medical relevant composite outcome, but preserve full statistical power. The test is also applicable in other situations where a specific value (say 0 days alive outside hospitals) encodes a special meaning. The test is implemented in an R package which is available for download.
KW - stat.ME
KW - stat.AP
M3 - Journal article
JO - arXiv.org: Statistics
JF - arXiv.org: Statistics
ER -
ID: 248695759