Prev: SQL equivalent of LAST. processing How to drop unwanted columns
Next: Restructure dataset
From: Thx for your HELP on 21 Jun 2010 13:58
Hi,

I am trying to fit an exponential curve on the following dataset.
The regression results I got from excel is: SpecHum = 4.7145 *
e^(0.0371*temperature)
How can I replicate the regression in SAS?
And also, is there a code that will automatically calculate SpecHum
for the temperature I want using the outputted regression line?

Any help is appreciated. Thank you very much.


data test;
input temperature SpecHum;
datalines;

105 145
100 145
100 140
100 130
95 140
95 130
95 130
95 120
95 110
90 140
90 130
90 120
90 110
90 100
90 90
85 140
85 130
85 120
85 110
85 100
85 90
85 80
80 140
80 130
80 120
80 110
80 100
80 90
80 80
80 70
75 110
75 100
75 90
75 80
75 70
75 60
70 100
70 90
70 80
70 70
70 60
70 50
65 80
65 70
65 60
65 50
65 40
60 70
60 60
60 50
60 40
55 50
55 40
55 30
50 40
50 30
45 40
45 30
45 20
40 30
40 20
35 20
30 20
25 10
25 15
20 10
20 8
15 10
15 8
10 8
10 6
5 6
5 4
0 4
-5 4
-10 3
-10 2
-15 2
-20 2
-20 1.5
;
run;
From: FPAStatman on 21 Jun 2010 14:44
On Jun 21, 11:58 am, Thx for your HELP <jjongw...(a)gmail.com> wrote:
> Hi,
>
> I am trying to fit an exponential curve on the following dataset.
> The regression results I got from excel is: SpecHum = 4.7145 *
> e^(0.0371*temperature)
> How can I replicate the regression in SAS?
> And also, is there a code that will automatically calculate SpecHum
> for the temperature I want using the outputted regression line?
>
> Any help is appreciated.  Thank you very much.
>
> data test;
>         input temperature SpecHum;
> datalines;
>
> 105     145
> 100     145
> 100     140
> 100     130
> 95      140
> 95      130
> 95      130
> 95      120
> 95      110
> 90      140
> 90      130
> 90      120
> 90      110
> 90      100
> 90      90
> 85      140
> 85      130
> 85      120
> 85      110
> 85      100
> 85      90
> 85      80
> 80      140
> 80      130
> 80      120
> 80      110
> 80      100
> 80      90
> 80      80
> 80      70
> 75      110
> 75      100
> 75      90
> 75      80
> 75      70
> 75      60
> 70      100
> 70      90
> 70      80
> 70      70
> 70      60
> 70      50
> 65      80
> 65      70
> 65      60
> 65      50
> 65      40
> 60      70
> 60      60
> 60      50
> 60      40
> 55      50
> 55      40
> 55      30
> 50      40
> 50      30
> 45      40
> 45      30
> 45      20
> 40      30
> 40      20
> 35      20
> 30      20
> 25      10
> 25      15
> 20      10
> 20      8
> 15      10
> 15      8
> 10      8
> 10      6
> 5       6
> 5       4
> 0       4
> -5      4
> -10     3
> -10     2
> -15     2
> -20     2
> -20     1.5
> ;
> run;

One way could be:

proc nlin data=test;
model spechum=b1*exp(b2*temperature);
parms b1=4
b2=0
;

run;

From: Thx for your HELP on 21 Jun 2010 15:12
Thanks for your help. I don't know if I am mistaken, but the code
doesn't seem to replicate the regression from excel. Is there another
option?

Thanks again.
From: FPAStatman on 21 Jun 2010 15:48
On Jun 21, 1:12 pm, Thx for your HELP <jjongw...(a)gmail.com> wrote:
> Thanks for your help.  I don't know if I am mistaken, but the code
> doesn't seem to replicate the regression from excel.  Is there another
> option?
>
> Thanks again.


How did you get the results from Excel? That might help get similar
results from SAS.
The model provided by Proc NLIN has smaller squared residuals than the
model you provided. (42,729.53 from your model, 20,411.68 from Proc
NLIN model). I also get the same results from Proc GENMOD, so more
details from Excel would be useful.
 | 
Pages: 1
Prev: SQL equivalent of LAST. processing How to drop unwanted columns
Next: Restructure dataset