Tableau calculated field using R to predict values using a polynomial fit of degree two. Data may contain zeroes which are excluded from the fit, but are predicted by the model.

polynomial.txt

SCRIPT_REAL("
    xdata <- .arg1
    ydata <- .arg2

    indices <- which( abs(ydata) < 0.01 )
    ndata <- length( ydata )
    weights <- rep(1, ndata)
    weights[indices] = 0

    model <- lm( ydata ~ xdata + I(xdata^2), weights=weights )
    model$fitted
",
SUM( [year] ), SUM( [13-17] ) )