Wondering if anyone can help with falls prediction. I am attempting to predict future falls on the basis of several collected physical measures (i.e. Step Test, Timed-Up-and-Go). I understand that looking at the occurrence of prior falls can be analyzed using regressions analysis, specifically logistic regression, but I am unsure how to go about predicting the likelihood of future falls. Any help in is greatly appreciated.
Try survival analysis " Kaplan-meier estimate".
You can use "0" as no fall and "1" as fall.
Agree with comment above, survival analysis and using the Kaplan-meier estimate
Article Understanding survival analysis: Kaplan-Meier estimate
I guess you should validate your logistic regression model for predictivity.
I should add that there is no follow-up period for the study. The future predictions are based purely on retrospective/cross-sectional data.
Not necessarily an issue, as long as you don't use the same data to estimate the model and validate the predictions (i.e. you should have a training dataset and a test dataset with no common observations).
Obviously, in this case you are making the (somehow strong) assumption that what you are studying through your model is stable in time.
I propose a kalman filter, applicable in all cases
Article Multi-site modeling and prediction of annual and monthly pre...
good luck
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Statistical Techniques