How to use contrast for phreg?

[SeaMoon_168]

The data have time, censor, and a categorical variable bleed (low, medium, high). I was asked to conduct a trend test for the bleed variable by adding

contrast 'linear' bleed -1 0 1;

after the model statement.

However, if the bleed has four categories, (low, medium, high, severe), the the contrast should be

contrast 'linear' bleed -3 -1 1 3 ;

Can anyone help me figure out why the values of the contrast have to be -2 1 1 2 instead or other values?

[Ksharp]

https://support.sas.com/kb/22/912.html

[SeaMoon_168]

Thank you for your help. But I am still confused about how these values are created, for example, linear (-3 -1 1 3) and quadratic(1 -1 -1 1)?

  contrast 'linear'    dose -3 -1  1 3;
  contrast 'quadratic' dose  1 -1 -1 1;
  contrast 'cubic'     dose -1  3 -3 1;

[Ksharp]

I think here CONSTRAST test the hypothesis:
H0: -3*does1 + -1*does2 + 1*does3 + 3*does4 =0
Once H0 is true, does1<does2<does3<does4

Here could have you some clue:
https://communities.sas.com/t5/Statistical-Procedures/Test-for-trend-using-median-values/m-p/306764

or calling @StatDave @SteveDenham @lvm @jiltao