Programming the statistical procedures from SAS

Question about two sample ttest

Accepted Solution Solved
Reply
New Contributor
Posts: 3
Accepted Solution

Question about two sample ttest

Dear scientists, I have a question about ttest.  

 

Description of my dataset: there are four colums. Colum GDM is the group 0 or 1. Y_m is the group mean. How can I carry out t-tests comparing GDM 0 to GDM 1 of each of the six outcomes which are SFA MUFA PUFA6 DHA PUFA3 N3_N6?

I proc my code but no result. Need help Smiley Happy

 

Here is the data

 

data fattyACID;
length ACID $5.;
input ACID $ GDM Y_m Y_sd;
if GDM=0 then Y_n=84; else Y_n=24;
datalines;
SFA 0 43.5 3.4
SFA 1 44.7 2.8
MUFA 0 24.5 2.2
MUFA 1 25.6 2.2
PUFA6 0 25.8 2.5
PUFA6 1 24.7 2.5
DHA 0 3.5 1.1
DHA 1 2.9 0.8
PUFA3 0 4.4 2.1
PUFA3 1 3.5 0.9
N3_N6 0 0.17 0.07
N3_N6 1 0.14 0.04
;
run;

Here is my proc, I think I follow the step which in SAS support , but I don't get result.(https://support.sas.com/documentation/cdl/en/statug/63033/HTML/default/viewer.htm#statug_ttest_sect0...)

 

proc sort;
   by acid gdm;
run;
proc print data=fattyacid;
run;

proc means data=fattyACID noprint;
class acid;
var y_m;
 by gdm;
 output out=newfattyACID;
run;
proc print data=newfattyACID;
run;

proc ttest data=newfattyACID;
by acid;
class gdm;
var y_m;
run; 

Accepted Solutions
Solution
‎03-20-2017 09:40 AM
Super User
Posts: 10,880

Re: Question about two sample ttest


Jackwangliang wrote:

Thank you dear balladw!

Then how about this data, I add y_n for the gdm group individual count, can SAS proc ttest. Or I still have to do it manually?

Thanks a lot!

 

data fattyACID;
length ACID $5.;
input ACID $ GDM Y_m Y_sd y_n;

 SFA 0 43.50 3.40 84 
 SFA 1 44.70 2.80 24 
 MUFA 0 24.50 2.20 84 
 MUFA 1 25.60 2.20 24 
 PUFA6 0 25.80 2.50 84 
 PUFA6 1 24.70 2.50 24 
 DHA 0 3.50 1.10 84 
 DHA 1 2.90 0.80 24 
 PUFA3 0 4.40 2.10 84 
 PUFA3 1 3.50 0.90 24 
 N3_N6 0 0.17 0.07 84 
 N3_N6 1 0.14 0.04 24 

;
run;

 

In the data prior to the Fattyacid data were all of the measurements exactly the same for each measurement? If so then you could use the above data with Freq y_n; But I doubt that was the case. Proc TTEST requires information to calculate the Standard Deviation within each group compared. That means some variability in the data . If you go look at your original Newfattyacid, assuming you managed to create it as you have syntax errors (fattyacid is sorted incorrectly) you will see that you have STD missing for EVERY acid GDM combination. If you can find a value for STD for each then you could use the data but Proc Means can't create one and you don't have any variability in the data attempting to use the FREQ option will either fail completely or yield very missleading results.

 

Also, if you do find the value for STD you will need to make the data set look like the output of proc means not the structure you currently have.

View solution in original post


All Replies
Super User
Posts: 10,880

Re: Question about two sample ttest

Since your data only includes one value for acid per gdm there isn't much to do here. The Proc TTEST expects to see one record per observation or have a variable that indicates the frequency count in the raw data for individual measurements.

 

If you have summary data, such as a mean, you would really need to have the n and standard deviation to manually to a ttest.

If you summarized another set to get your Fattyacid data use that one for proc ttest.

New Contributor
Posts: 3

Re: Question about two sample ttest

Thank you dear balladw!

Then how about this data, I add y_n for the gdm group individual count, can SAS proc ttest. Or I still have to do it manually?

Thanks a lot!

 

data fattyACID;
length ACID $5.;
input ACID $ GDM Y_m Y_sd y_n;

 SFA 0 43.50 3.40 84 
 SFA 1 44.70 2.80 24 
 MUFA 0 24.50 2.20 84 
 MUFA 1 25.60 2.20 24 
 PUFA6 0 25.80 2.50 84 
 PUFA6 1 24.70 2.50 24 
 DHA 0 3.50 1.10 84 
 DHA 1 2.90 0.80 24 
 PUFA3 0 4.40 2.10 84 
 PUFA3 1 3.50 0.90 24 
 N3_N6 0 0.17 0.07 84 
 N3_N6 1 0.14 0.04 24 

;
run;
Solution
‎03-20-2017 09:40 AM
Super User
Posts: 10,880

Re: Question about two sample ttest


Jackwangliang wrote:

Thank you dear balladw!

Then how about this data, I add y_n for the gdm group individual count, can SAS proc ttest. Or I still have to do it manually?

Thanks a lot!

 

data fattyACID;
length ACID $5.;
input ACID $ GDM Y_m Y_sd y_n;

 SFA 0 43.50 3.40 84 
 SFA 1 44.70 2.80 24 
 MUFA 0 24.50 2.20 84 
 MUFA 1 25.60 2.20 24 
 PUFA6 0 25.80 2.50 84 
 PUFA6 1 24.70 2.50 24 
 DHA 0 3.50 1.10 84 
 DHA 1 2.90 0.80 24 
 PUFA3 0 4.40 2.10 84 
 PUFA3 1 3.50 0.90 24 
 N3_N6 0 0.17 0.07 84 
 N3_N6 1 0.14 0.04 24 

;
run;

 

In the data prior to the Fattyacid data were all of the measurements exactly the same for each measurement? If so then you could use the above data with Freq y_n; But I doubt that was the case. Proc TTEST requires information to calculate the Standard Deviation within each group compared. That means some variability in the data . If you go look at your original Newfattyacid, assuming you managed to create it as you have syntax errors (fattyacid is sorted incorrectly) you will see that you have STD missing for EVERY acid GDM combination. If you can find a value for STD for each then you could use the data but Proc Means can't create one and you don't have any variability in the data attempting to use the FREQ option will either fail completely or yield very missleading results.

 

Also, if you do find the value for STD you will need to make the data set look like the output of proc means not the structure you currently have.

New Contributor
Posts: 3

Re: Question about two sample ttest

Thank you! This is the point.

☑ This topic is solved.

Need further help from the community? Please ask a new question.

Discussion stats
  • 4 replies
  • 122 views
  • 0 likes
  • 2 in conversation