I built a prediction model following a linear regression as :

Mod <- lm(Rate ~ OM + SOC + ST + Cover, data=Data)

enter image description here

On the graph (The black line is the regression X= Observed VS Y= Fitted values,the yellow line is the bissector Y=X), you can see that my model underestimate the Rate when > 5 and overestimate it when <5.

In order to decrease the error when I use my model to predict, is it correct to establish the quantile regressions as follow :

QR25 <- rq(Rate ~ OM + SOC + ST + Cover,tau =0.25, data=Data)
QR75 <- rq(Rate ~ OM + SOC + ST + Cover,tau =0.75, data=Data)

and to use QR25 to predict the values when Mod predict Rate < 5 and QR75 when Mod predict Rate >5 ?

On the second graph, the blue line is the trend line for the points predicted by QR25 and the orange line is the trend line for the points predicted by QR0.75enter image description here

Thank you in advance !