topic Re: Errant Graph when ploting predicted values from proc nlin in Graphics Programming

Errant graph when plotting predicted values from proc nlin

RobinN — Thu, 29 Oct 2020 17:37:49 GMT

Hello,

I'm currently working on a SAS Programm to fit my data of a germinating fungus to a given logistic function. This works fine for now, as the estimated values fit well to the data and predicted values based on that function seem acceptable. But when I try to plot this data the resulting graph looks like this:

Does anyone has any suggestions regarding this problem? I'm using SAS University Edition / SAS Studio 3.8

Thanks for your help!

The code used to create this graph is as follows:

data fungi;
     input time germrate;
     datalines;
9       0
9       0
9       0
9       0
9       0
9       0
9       0
9       0
12      1
12      1
12      1
9       1
12      2
12      3
12      4
12      4
12      5
12      5
15      11
15      18
15      20
15      21
9       22
15      24
15      27
15      27
9       28
9       28
9       29
9       31
9       33
15      34
15      37
18      43
18      45
18      52
18      52
18      54
15      55
18      58
18      58
18      63
18      71
21      72
12      73
21      74
12      74
12      76
12      76
12      77
21      78
15      79
12      80
21      81
15      82
21      83
21      84
21      84
24      85
15      85
24      86
21      86
24      87
21      87
21      87
24      88
24      90
24      90
24      90
21      90
18      90
18      91
15      91
24      92
24      92
21      92
21      92
21      92
18      92
24      93
24      93
18      93
15      93
24      94
24      94
24      95
24      96
18      96
21      97
18      98
96      99
96      100
96      99
96      98
96      97
96      100

   ;


proc nlin  data=fungi;
 	parms   Pmax=98 teta=15 ;d=6 ;

 	model germrate = Pmax*(1-(1/(1+((time/teta)**d))));
run;

proc nlin data=fungi method=newton;   
	parameters  Pmax=98 teta=15 ;d=6 ;  
	model germrate = Pmax*(1-(1/(1+((time/teta)**d)))); 
	output out=nlinout predicted=pred l95m=l95mean u95m=u95mean l95=l95ind u95=u95ind; 
run; 
proc print data=nlinout; 
run;

data filler;   
do time = 0 to 96 by 2;     
predict=1;     
output;   
end; 
run;

data fitthis; 
set fungi filler; 
run;

proc nlin data=fitthis method=newton;   
	parameters  Pmax=98 teta=15 ;d=6 ;  
	model germrate = Pmax*(1-(1/(1+((time/teta)**d))));   
	output out=nlinout predicted=pred l95m=l95mean u95m=u95mean; 
run; 

proc print data=nlinout(where=(predict=1)); run;

proc sgplot data=nlinout;

   scatter x=time y=germrate;
   series x=time y=pred;
   band upper=u95mean lower=l95mean x=time ;
   
   run;

Re: Errant Graph when ploting predicted values from proc nlin

djrisks — Thu, 29 Oct 2020 17:36:44 GMT

Hi @RobinN , you just need to sort your graph by time before you do the plot. Please see the example below.

data fungi;
     input time germrate;
     datalines;
9       0
9       0
9       0
9       0
9       0
9       0
9       0
9       0
12      1
12      1
12      1
9       1
12      2
12      3
12      4
12      4
12      5
12      5
15      11
15      18
15      20
15      21
9       22
15      24
15      27
15      27
9       28
9       28
9       29
9       31
9       33
15      34
15      37
18      43
18      45
18      52
18      52
18      54
15      55
18      58
18      58
18      63
18      71
21      72
12      73
21      74
12      74
12      76
12      76
12      77
21      78
15      79
12      80
21      81
15      82
21      83
21      84
21      84
24      85
15      85
24      86
21      86
24      87
21      87
21      87
24      88
24      90
24      90
24      90
21      90
18      90
18      91
15      91
24      92
24      92
21      92
21      92
21      92
18      92
24      93
24      93
18      93
15      93
24      94
24      94
24      95
24      96
18      96
21      97
18      98
96      99
96      100
96      99
96      98
96      97
96      100

   ;


proc nlin  data=fungi;
 	parms   Pmax=98 teta=15 ;d=6 ;

 	model germrate = Pmax*(1-(1/(1+((time/teta)**d))));
run;

proc nlin data=fungi method=newton;   
	parameters  Pmax=98 teta=15 ;d=6 ;  
	model germrate = Pmax*(1-(1/(1+((time/teta)**d)))); 
	output out=nlinout predicted=pred l95m=l95mean u95m=u95mean l95=l95ind u95=u95ind; 
run; 
proc print data=nlinout; 
run;

data filler;   
do time = 0 to 96 by 2;     
predict=1;     
output;   
end; 
run;

data fitthis; 
set fungi filler; 
run;

proc nlin data=fitthis method=newton;   
	parameters  Pmax=98 teta=15 ;d=6 ;  
	model germrate = Pmax*(1-(1/(1+((time/teta)**d))));   
	output out=nlinout predicted=pred l95m=l95mean u95m=u95mean; 
run; 

proc print data=nlinout(where=(predict=1)); run;

proc sort data = nlinout;
  by time;
run;

proc sgplot data=nlinout;

   scatter x=time y=germrate;
   series x=time y=pred;
   band upper=u95mean lower=l95mean x=time ;
run;

Re: Errant Graph when ploting predicted values from proc nlin

djrisks — Thu, 29 Oct 2020 17:39:18 GMT

I meant to say, sort your data instead of sort your graph...

Re: Errant Graph when ploting predicted values from proc nlin

RobinN — Thu, 29 Oct 2020 17:42:54 GMT

This was quite easy and obvious. Many thanks for your quick help!

Re: Errant Graph when ploting predicted values from proc nlin

djrisks — Thu, 29 Oct 2020 17:45:57 GMT

You're welcome! When you're doing series and band plots it's best to sort your data beforehand by the variable on the x-axis 🙂