LMF reports the same predictions for newdata

[shengzhuo-yuan]

When I use the LMF to predict the slopes hk(x) for new data, the prediction is the same for all samples, which should differ for different x.

[shengzhuo-yuan]

#train linear model forest
lmf <- lm_forest(iv_x.s16, sch_exp.s16, iv_w.s16)
#prediction for section 16
tau.hat.s16 <- predict(lmf)$predictions
#make it 2D instead of 3D
tau.hat.s16 <- drop(tau.hat.s16)
#prediction for non section 16
tau.hat.ns16 <- predict(lmf, newdata=iv_x.ns16)$predictions
tau.hat.ns16 <- drop(tau.hat.ns16)

The predictions for non section 16 are the same for all observations though they have different x

[erikcs]

Hi @shengzhuo-yuan, if, for example, your trained lm forest did not split on any covariates, the predictions will all be the same.