Plot spline based coefficients from a coxph model from the survival package in R

termplot_coxph.R

# based on https://cran.r-project.org/web/packages/survival/vignettes/splines.pdf from Terry Therneau

# start with termplot without the plot to return results for all coefficients in the model

# y is the object in which the coxph model results have been saved

d <- termplot(y, se = TRUE, plot = FALSE)


# takes the termplot object (tp_obj), a specific variable name from the model as a string (var_name), and outputs a plot
# allows user to pass an x-axis label and to restrict the input range (e.g., input_range = c(10,30) will restrict the values 
# to 10-30 inclusive).  Need to be careful that the range makes sense.

# base plot version
spline_plot <- function(tp_obj, var_name = "", var_label = var_name, input_range = NULL){
    dat <- tp_obj[[var_name]]
    if(!is.null(input_range)){
        dat <- dat[dat$x >= input_range[1] & dat$x <= input_range[2], ]
    }
    matplot(
        dat$x, 
        exp(dat$y + outer(dat$se, c(0, -1.96, 1.96), '*')), 
        log ='y', 
        type ='l', 
        lty = c(1,2,2), 
        col = 1, 
        xlab = var_label, 
        ylab = "Relative death rate")
}

# ggplot2 version
# edit scale_y_continuous breaks to set different y-axis values to be shown
# edit the geom color settings for different colors
# add title, subtitle, and/or caption to labs as needed
spline_plot_gg <- function(tp_obj, var_name = "", var_label = var_name, input_range = NULL){
    dat <- tp_obj[[var_name]]
    if(!is.null(input_range)){
        dat <- dat[dat$x >= input_range[1] & dat$x <= input_range[2], ]
    }
    ggplot(dat, aes(x, exp(y))) +
        geom_ribbon(aes(ymin = exp(y - 1.96 * se), ymax = exp(y + 1.96 * se)), fill = "lightgrey") + 
        geom_line(color = "black") +
        scale_y_continuous(trans = "log", breaks = c(0.2, 0.5, 1, 1.5, 2, 3)) +
        theme_bw() +
        labs(
            x = var_label,
            y = "Relative death rate")
}

Please feel free to use/adapt the above for any purpose. Thanks to Terry Therneau for providing the instructions in his vignette.