Table 3.
Summary of methods and performance. The first three columns show what data features the method accounts for by design. The last three columns summarize relative performance of the methods and indicates the recommended method based on study objectives of estimating the exposure-response function (ER) and detecting interactions among exposures. These recommendations are based on the simulation study results.
Method | Method accounts for by design |
Method recommended use |
||||
---|---|---|---|---|---|---|
exposure timing |
inter- actions |
non- linearity |
detecting interactions |
Estimating ER | ||
wk signal* modest n |
st signal* large n |
|||||
Add. DLM | ✓ | |||||
Add. DLNM | ✓ | ✓ | ||||
BKMR | ✓ | ✓ | ✓ | |||
BKMR-DLM | ✓ | ✓ | ✓ | ✓ | ✓ |
denotes weaker signal and stronger signal, respectively.
ER: exposure-response