What about this code is not working for you?   data test; input Forest_Type $ X Y; datalines; CO 16.8 16.8 CO 16 16.5 CO 16.5 16.9 CO 17 17.5 CO 18.2 18.2 CO 17.8 17.9 OP 17.9 18.1 OP 17.9 16.7 OP 13.9 14.9 OP 16.7 16.2 OP 17.8 17.8 OP 17 17.9 OP 17.2 16.8 OP 17.5 17.2 OP 17.9 16.7 PO 15.9 15.8 PO 14.4 13.4 PO 17.6 17.6 PO 15.2 15.8 PO 15.8 19.3 PO 16 16 PO 16.1 16 PP 16.4 15.6 PP 16.4 17.1 PP 16.9 16.6 PP 15.2 15.9 PP 16.6 16.4 SO 13.8 14.4 SO 14.9 14.6 SO 15.9 15.6 SO 15.8 19 ; run; proc sgplot data=test; scatter x=x y=y / group=forest_type; reg x=x y=y / degree=1; reg x=x y=y / degree=2; lineparm x=0 y=0 slope=1 / clip; run; Here is the graph produced by this code:   ... View more

That is acceptable.   If "canopy" measures either a fixed area of shade generating plant material or percentage of ground covered by such then one would expect a reduction of temperature for a given solar radiation input based on the shade.   Hint: a model without the solar radiation input may generate some odd results because the "canopy" likely doesn't change quickly but temperature can based on the sunshine. (Actually worked a project with ground temperature sensors north, east, west and south of clumps of bunch grass and dealt with such to model ground temperature.)   Remember the underlying generic linear function is a y= mx+b; m is the slope, b an intercept. There is no restriction in general on whether X is positive, negative or 0.   IF a value you measure should not negative values, such as Relative Humidity, then interpretation of a model may require restriction if the dependent value goes negative. ... View more