topic Re: Why i had 2 p-value in nonlinearity test of restricted cubic spline in New SAS User

Why i had 2 p-value in nonlinearity test of restricted cubic spline

isozealot — Fri, 01 Nov 2024 14:44:06 GMT

Dear all:

I used restricted cubic spline (RCS) to plot the mortality hazard ratio of body mass index. The hazard ratio was adjusted by age, viral type, SOFA score, and bicarbionate.

However, there were two results of P-value for nonlinearity (Counts1 = 0.2264 and Counts2 = 0.6636).

Which one should i use?

I used BMI=18.5 as the single knot. The following figure was the RCS plot.

Re: Why i had 2 p-value in nonlinearity test of restricted cubic spline

ballardw — Fri, 01 Nov 2024 14:54:20 GMT

Please include the code you submitted.

That way we have a chance of seeing all of the options that could have affected the creation of the output.

Re: Why i had 2 p-value in nonlinearity test of restricted cubic spline

Rick_SAS — Sun, 03 Nov 2024 10:25:18 GMT

If you are getting 2 estimates, it means that you are using 3 knots for the splines. The procedure probably outputs a "SplineKnots" table, which will show the 3 values of the continuous variable at which the knots are placed. For example, see this example:

proc glmselect data=sashelp.cars;
   effect CONUTs = spline(Weight / details naturalcubic basis=tpf(noint)
             knotmethod=equal(3)); /* <== NOTE: 3 equally spaced knots */
   model mpg_city = CONUTs / selection=none;       /* fit model by using spline effects */
   ods select SplineKnots ParameterEstimates;
quit;

The three spline components are the Intercept, CONUTs_1, and CONUTS_2, The p-values indicate whether each component in the model is a significant predictor of the response variable. For more about the spline components (and how to visualize them), see https://blogs.sas.com/content/iml/2019/10/16/visualize-regression-splines.html