Dataset Information

Impact of correlation on predictive ability of biomarkers.

ABSTRACT: In this paper, we investigate how the correlation structure of independent variables affects the discrimination of risk prediction model. Using multivariate normal data and binary outcome, we prove that zero correlation among predictors is often detrimental for discrimination in a risk prediction model and negatively correlated predictors with positive effect sizes are beneficial. A very high multiple R-squared from regressing the new predictor on the old ones can also be beneficial. As a practical guide to new variable selection, we recommend to select predictors that have negative correlation with the risk score based on the existing variables. This step is easy to implement even when the number of new predictors is large. We illustrate our results by using real-life Framingham data suggesting that the conclusions hold outside of normality. The findings presented in this paper might be useful for preliminary selection of potentially important predictors, especially is situations where the number of predictors is large.

SUBMITTER: Demler OV

PROVIDER: S-EPMC4177016 | biostudies-literature | 2013 Oct

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Impact of correlation on predictive ability of biomarkers.

Demler Olga V OV Pencina Michael J MJ D'Agostino Ralph B RB

Statistics in medicine 20130503 24

In this paper, we investigate how the correlation structure of independent variables affects the discrimination of risk prediction model. Using multivariate normal data and binary outcome, we prove that zero correlation among predictors is often detrimental for discrimination in a risk prediction model and negatively correlated predictors with positive effect sizes are beneficial. A very high multiple R-squared from regressing the new predictor on the old ones can also be beneficial. As a practi ...[more]

PMID: 23640729

Similar Datasets

Project description:BackgroundA longstanding gap in the reproductive health field has been the availability of a screening instrument that can reliably predict a person's likelihood of becoming pregnant. The Desire to Avoid Pregnancy Scale is a new measure; understanding its sensitivity and specificity as a screening tool for pregnancy as well as its predictive ability and how this varies by socio-demographic factors is important to inform its implementation.MethodsThis analysis was conducted on a cohort of 994 non-pregnant participants recruited in October 2018 and followed up for one year. The cohort was recruited using social media as well as advertisements in a university, school, abortion clinic and outreach sexual health service. Almost 90% of eligible participants completed follow-up at 12 months; those lost to follow-up were not significantly different on key socio-demographic factors. We used baseline DAP score and a binary variable of whether participants experienced pregnancy during the study to assess the sensitivity, specificity, area under the ROC curve (AUROC) and positive and negative predictive values (PPV and NPV) of the DAP at a range of cut-points. We also examined how the predictive ability of the DAP varied according to socio-demographic factors and by the time frame considered (e.g., pregnancy within 3, 6, 9 and 12 months).ResultsAt a cut-point of 2 on the 0-4 range of the DAP scale, the DAP had a sensitivity of 0.78, a specificity of 0.81 and an excellent AUROC of 0.87. In this sample the cumulative incidence of pregnancy was 16% (95%CI 13%, 18%) making the PPV 43% and the NPV 95% at this cut-point. The DAP score was the factor most strongly associated with pregnancy, even after age and number of children were taken into account. The association between baseline DAP score and pregnancy did not differ across time frames.ConclusionsThis is the first study to assess the DAP scale as a screening tool and shows that its predictive ability is superior to the limited pre-existing pregnancy prediction tools. Based on our findings, the DAP could be used with a cut-point selected according to the purpose.

Dataset Information

Impact of correlation on predictive ability of biomarkers.

Publications

Impact of correlation on predictive ability of biomarkers.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets