“…Here, a penalized splines method is adopted to approximate g ( t ). Following Lang and Brezger 6 and Tang et al, 10 the polynomial splines approximation of g ( t ) has the form where q is the degree of the polynomial component, K is the number of knots ( K knots divide the possible values of g ( t ) into K + 1 regression intervals), is a ( q + K )‐vector of unknown coefficients, and with , and is the location of the k th knot and is usually set as the {( k + 1)/( K + 2)}th quantile of the unique dataset { t i : i = 1, … , n } for k = 1, … , K or equidistantly set in the interval (…”