#Sas robustreg fitplot how to

The vertical bar creates the interaction between the CLASS variable and the quantitative variable. The ZERO=NONE option creates one binary variable for each CLASS level and suppresses the overall intercept (hence, the model has an implicit intercept). This PROC TRANSREG step fits a model with three intercepts and three slopes (one each for each of the three iris species). In the second part of this post, I will show how you can use PROC TRANSREG to fit a model where the fit function for each group is not independent of the other groups. However, I quite deliberately built these kinds of models into the procedure from its earliest days. I wrote PROC TRANSREG, so I no doubt turn to it more than most people. Rick used PROC REG and a BY statement to get the parameter estimates for the separate fit functions. This post also uses discrete attribute maps.

#Sas robustreg fitplot how to

Since we do not have gridded layouts in PROC SGPLOT, I'll show how to use an axis table instead. When PROC writers are faced with creating a table like the ones that you see in this post, they use the GTL and a LAYOUT GRIDDED. Usually, you use axis tables when there is a clear link between the rows of the axis table and the graph. This is an unusual illustration of axis tables. In contrast, a table can handle more groups (since no attempt is made to align the equation and the parameter estimates) and you can additionally add other statistics such t statistics and p values.

Rick's labels clearly show the association between the function and its equation. Rick suggested that I might want to approach the same topic but by using axis tables. Each might have advantages and disadvantages. There are often several approaches that one can take to a problem. Astute readers of the Graphically Speaking and DO Loop blogs might have noticed that Rick and I sometimes tag-team on topics, with each of us putting our own spin on them. A recent post in Rick Wicklin's DO Loop blog shows how to label multiple regression lines.