TY - GEN
T1 - One Patient, Two Patient, Three Patient, Four - When Patients Are Counted, but Not Accounted For
T2 - 9th International Conference on Information Technology in Medicine and Education, ITME 2018
AU - Manno, Francis A.M.
AU - Manno, Sinai H.C.
AU - Ahmed, Irfan
AU - Liu, Yuanchao
AU - Cheng, Shuk Han
AU - Lau, Condon
N1 - Publisher Copyright:
© 2018 IEEE.
PY - 2018/12/26
Y1 - 2018/12/26
N2 - Background: Pseudoreplication is a statistical error occurring when treatments are not replicated or replicates are not statistically independent. Description and explanation of pseudoreplication is absent from medical literature. There are four types of pseudoreplication: simple, temporal, implicit, and pooling. Methods: The present article reviews the occurrence of type I errors caused by pseudoreplication in prominent medical journals. Fifty articles, ten randomly chosen from five prominent journals: JAMA, BMJ, The New England Journal of Medicine, The Lancet, and The Annals of Internal Medicine, were analyzed for the occurrence of pseudoreplication, and in particular which statistical comparison was utilized most resulting in this error. The unit of analysis was the article for group assessments and the statistical test within each article for comparisons. Multivariate analyses were carried out to determine the correlation of pseudoreplication type with parameters in the manuscript. Results: A total of 1,294 statistical comparisons of different hypotheses were assessed, averaging 25.9 (SE=4.2) per article. Pseudoreplication occurred an average of 37.38% (SE=5.0) per statistic utilized in studies. Temporal pseudoreplication occurred most frequently. Temporal and implicit pseudoreplication were significantly related to the first statistic (F=3.45, df=6,43, P<0.007 and F=5.32, df=6,43, P<0.0004, respectively) and simple pseudoreplication was significantly related to the second statistic utilized, F=2.81, df=5,44, P<0.03. The number of M.D.'s on a study was significantly related to pseudoreplication as determined by Roy's Largest Root = 1.2, F=2.18, df=13,23, P<0.05. Conclusions: Pseudoreplication is a serious statistical problem related to the improper implementation and analysis of statistics. Physicians and medical scientists need to be cognizant of the manner in which the error occurs to prevent Type I error.
AB - Background: Pseudoreplication is a statistical error occurring when treatments are not replicated or replicates are not statistically independent. Description and explanation of pseudoreplication is absent from medical literature. There are four types of pseudoreplication: simple, temporal, implicit, and pooling. Methods: The present article reviews the occurrence of type I errors caused by pseudoreplication in prominent medical journals. Fifty articles, ten randomly chosen from five prominent journals: JAMA, BMJ, The New England Journal of Medicine, The Lancet, and The Annals of Internal Medicine, were analyzed for the occurrence of pseudoreplication, and in particular which statistical comparison was utilized most resulting in this error. The unit of analysis was the article for group assessments and the statistical test within each article for comparisons. Multivariate analyses were carried out to determine the correlation of pseudoreplication type with parameters in the manuscript. Results: A total of 1,294 statistical comparisons of different hypotheses were assessed, averaging 25.9 (SE=4.2) per article. Pseudoreplication occurred an average of 37.38% (SE=5.0) per statistic utilized in studies. Temporal pseudoreplication occurred most frequently. Temporal and implicit pseudoreplication were significantly related to the first statistic (F=3.45, df=6,43, P<0.007 and F=5.32, df=6,43, P<0.0004, respectively) and simple pseudoreplication was significantly related to the second statistic utilized, F=2.81, df=5,44, P<0.03. The number of M.D.'s on a study was significantly related to pseudoreplication as determined by Roy's Largest Root = 1.2, F=2.18, df=13,23, P<0.05. Conclusions: Pseudoreplication is a serious statistical problem related to the improper implementation and analysis of statistics. Physicians and medical scientists need to be cognizant of the manner in which the error occurs to prevent Type I error.
KW - Assumptions
KW - Errors
KW - Pseudoreplication
KW - Replicates
KW - Samples
KW - Unit of Analysis
UR - https://www.scopus.com/pages/publications/85061341576
U2 - 10.1109/ITME.2018.00065
DO - 10.1109/ITME.2018.00065
M3 - Conference contribution
AN - SCOPUS:85061341576
T3 - Proceedings - 9th International Conference on Information Technology in Medicine and Education, ITME 2018
SP - 260
EP - 264
BT - Proceedings - 9th International Conference on Information Technology in Medicine and Education, ITME 2018
Y2 - 19 October 2018 through 21 October 2018
ER -