For the current SHAP value for the survival analysis, the outcome is with the right censoring format and covariates measured at one time point for each patient; And I am wondering if the SHAP value for the survival analysis can be generated with time-varying covariates for the same ID?
For example, In the coxph function, if the covariate only measured at one timepoint before the event, the surv object will be specified with Surv(time, status); With time-varying covariates, one id might have several rows (1...j) and the Surv object will be specified in the format like Surv(start, stop, status) with the time-varying covariates in the counting process format. The status will always be 0 in 1st to j-1row and will be 0/1 in the jth row.
Thanks!
For the current SHAP value for the survival analysis, the outcome is with the right censoring format and covariates measured at one time point for each patient; And I am wondering if the SHAP value for the survival analysis can be generated with time-varying covariates for the same ID?
For example, In the coxph function, if the covariate only measured at one timepoint before the event, the surv object will be specified with Surv(time, status); With time-varying covariates, one id might have several rows (1...j) and the Surv object will be specified in the format like Surv(start, stop, status) with the time-varying covariates in the counting process format. The status will always be 0 in 1st to j-1row and will be 0/1 in the jth row.
Thanks!