I am a very new R user. And for a project I need to estimate a “Semiparametric smooth-coefficient (SPSC)’ model following this paper (here).

I have cross-section data on 1,000 households. The dependent variable is Gross_returns, which is a function of land and labor inputs. In addition, I have 3 risk variables (Risk1, Risk2, Risk3) that enter the production function in some unspecified nonlinear way. My model is
Returns = α(Z) + Landβ1*(Z) + Laborβ2(Z) + ui
Going though an earlier post (here) I understand that I need to use something like

library(tidyverse)
library(readxl) 

Data=read_excel("C:/Users/A/Desktop/Data.xlsx")
NP.Ydata <- Data$Ln_Gross_ret 
NP.Xdata <- Data[c("Ln_Land", "Ln_Labor")] 
NP.Zdata <- Data[c("risk_behavior", "risk_pref", "agt_exper")] 

NP.bw.scoef <- npscoefbw(xdat=NP.Xdata, ydat=NP.Ydata, zdat=NP.Zdata)
NP.scoef <- npscoef(NP.bw.scoef, 
                    betas = TRUE,  
                    residuals = TRUE,
                    errors = TRUE)

I am able to run the above model. However, I have the following questions.
I need the direct, indirect and overall marginal impacts of the risk variables.
I need to add constraints to make the marginal products (elasticities) non-negative. My reference paper says that it uses “quadprog” for this.

Any suggestions for the above questions would be appreciated.