S.E. & C.D. in Split Plot design with SAS
in [SAS]
|
From: ManojK on 6 Mar 2010 00:51 Hi all, I want to calculate the following Standard Errors and Crirtical Differences for Split Plot design using SAS. The formulas as follows: 1. Estimate of S.E. of difference b/w two main plot treatment means = Sqrt(2*E1/r*b) 2. Estimate of S.E. of difference b/w two sub plot treatment means = Sqrt(2*E2/r*a) 3. Estimate of S.E. of difference b/w two sub plot treatment means at the same level of main plot treatment = Sqrt(2*E2/r) 4. Estimate of S.E. of difference b/w two main plot treatment means at the same or different levels of sub plot treatment = Sqrt((2*[(b-1)*E2 + E1])/ r*b) Critical difference is obtained by multiplying the S.E by t5% table value for respective error d.f. for 1,2,3. For 4, as the standard error of mean difference involves two error terms, we use the following equation to compute the weighted t values: t = [(b-1)*E2*ta + E1*ta] / [(b-1)*E2 + E1] where r = no. of replications a = no. of main plots b = no. of sub plots E1 = MSE of main plot error (obtained in the analysis using SAS) E2 = MSE of sub plot error (obtained in the analysis using SAS) ta = t-value at error d.f. (E1) tb = t-value at error d.f. (E2) All details are given above. Now please tell me how I can calculate these things using SAS? Thanks in anticipation. Manoj Khandelwal SRF, IASRI Delhi
From: ManojK on 6 Mar 2010 00:57 On Mar 6, 10:51 am, ManojK <khandelwalmano...(a)gmail.com> wrote: > Hi all, > > I want to calculate the following Standard Errors and Crirtical > Differences for Split Plot design using SAS. The formulas as follows: > > 1. Estimate of S.E. of difference b/w two main plot treatment means = > Sqrt(2*E1/r*b) > > 2. Estimate of S.E. of difference b/w two sub plot treatment means = > Sqrt(2*E2/r*a) > > 3. Estimate of S.E. of difference b/w two sub plot treatment means at > the same level of main plot treatment = Sqrt(2*E2/r) > > 4. Estimate of S.E. of difference b/w two main plot treatment means at > the same or different levels of sub plot treatment = Sqrt((2*[(b-1)*E2 > + E1])/ r*b) > > Critical difference is obtained by multiplying the S.E by t5% table > value for respective error d.f. for 1,2,3. For 4, as the standard > error of mean difference involves two error terms, we use the > following equation to compute the weighted t values: > > t = [(b-1)*E2*ta + E1*ta] / [(b-1)*E2 + E1] > > where r = no. of replications > a = no. of main plots > b = no. of sub plots > E1 = MSE of main plot error (obtained in the analysis using > SAS) > E2 = MSE of sub plot error (obtained in the analysis using > SAS) > ta = t-value at error d.f. (E1) > tb = t-value at error d.f. (E2) > > All details are given above. Now please tell me how I can calculate > these things using SAS? > > Thanks in anticipation. > > Manoj Khandelwal > SRF, IASRI > Delhi The data and code is given here: data split; input rep A B yield; cards; 1 1 1 25.0 1 1 2 31.0 1 2 1 21.5 1 2 2 24.0 2 1 1 41.2 2 1 2 36.1 2 2 1 40.4 2 2 2 37.8 3 1 1 26.3 3 1 2 32.4 3 2 1 28.6 3 2 2 27.6 ; proc glm; class rep A B; model yield= rep A rep*A B A*B; test h=A e=rep*A; run
From: data _null_; on 7 Mar 2010 08:33 On Mar 5, 11:57 pm, ManojK <khandelwalmano...(a)gmail.com> wrote: > On Mar 6, 10:51 am, ManojK <khandelwalmano...(a)gmail.com> wrote: > > > > > > > Hi all, > > > I want to calculate the following Standard Errors and Crirtical > > Differences for Split Plot design using SAS. The formulas as follows: > > > 1. Estimate of S.E. of difference b/w two main plot treatment means = > > Sqrt(2*E1/r*b) > > > 2. Estimate of S.E. of difference b/w two sub plot treatment means = > > Sqrt(2*E2/r*a) > > > 3. Estimate of S.E. of difference b/w two sub plot treatment means at > > the same level of main plot treatment = Sqrt(2*E2/r) > > > 4. Estimate of S.E. of difference b/w two main plot treatment means at > > the same or different levels of sub plot treatment = Sqrt((2*[(b-1)*E2 > > + E1])/ r*b) > > > Critical difference is obtained by multiplying the S.E by t5% table > > value for respective error d.f. for 1,2,3. For 4, as the standard > > error of mean difference involves two error terms, we use the > > following equation to compute the weighted t values: > > > t = [(b-1)*E2*ta + E1*ta] / [(b-1)*E2 + E1] > > > where r = no. of replications > > a = no. of main plots > > b = no. of sub plots > > E1 = MSE of main plot error (obtained in the analysis using > > SAS) > > E2 = MSE of sub plot error (obtained in the analysis using > > SAS) > > ta = t-value at error d.f. (E1) > > tb = t-value at error d.f. (E2) > > > All details are given above. Now please tell me how I can calculate > > these things using SAS? > > > Thanks in anticipation. > > > Manoj Khandelwal > > SRF, IASRI > > Delhi > > The data and code is given here: > > data split; > input rep A B yield; > cards; > 1 1 1 25.0 > 1 1 2 31.0 > 1 2 1 21.5 > 1 2 2 24.0 > 2 1 1 41.2 > 2 1 2 36.1 > 2 2 1 40.4 > 2 2 2 37.8 > 3 1 1 26.3 > 3 1 2 32.4 > 3 2 1 28.6 > 3 2 2 27.6 > ; > > proc glm; > class rep A B; > model yield= rep A rep*A B A*B; > test h=A e=rep*A; > run- Hide quoted text - > > - Show quoted text - This may help http://www.ats.ucla.edu/stat/SAS/library/SASExpDes_os.htm
From: ManojK on 7 Mar 2010 08:47 On Mar 7, 6:33 pm, "data _null_;" <datan...(a)gmail.com> wrote: > On Mar 5, 11:57 pm, ManojK <khandelwalmano...(a)gmail.com> wrote: > > > > > On Mar 6, 10:51 am, ManojK <khandelwalmano...(a)gmail.com> wrote: > > > > Hi all, > > > > I want to calculate the following Standard Errors and Crirtical > > > Differences for Split Plot design using SAS. The formulas as follows: > > > > 1. Estimate of S.E. of difference b/w two main plot treatment means = > > > Sqrt(2*E1/r*b) > > > > 2. Estimate of S.E. of difference b/w two sub plot treatment means = > > > Sqrt(2*E2/r*a) > > > > 3. Estimate of S.E. of difference b/w two sub plot treatment means at > > > the same level of main plot treatment = Sqrt(2*E2/r) > > > > 4. Estimate of S.E. of difference b/w two main plot treatment means at > > > the same or different levels of sub plot treatment = Sqrt((2*[(b-1)*E2 > > > + E1])/ r*b) > > > > Critical difference is obtained by multiplying the S.E by t5% table > > > value for respective error d.f. for 1,2,3. For 4, as the standard > > > error of mean difference involves two error terms, we use the > > > following equation to compute the weighted t values: > > > > t = [(b-1)*E2*ta + E1*ta] / [(b-1)*E2 + E1] > > > > where r = no. of replications > > > a = no. of main plots > > > b = no. of sub plots > > > E1 = MSE of main plot error (obtained in the analysis using > > > SAS) > > > E2 = MSE of sub plot error (obtained in the analysis using > > > SAS) > > > ta = t-value at error d.f. (E1) > > > tb = t-value at error d.f. (E2) > > > > All details are given above. Now please tell me how I can calculate > > > these things using SAS? > > > > Thanks in anticipation. > > > > Manoj Khandelwal > > > SRF, IASRI > > > Delhi > > > The data and code is given here: > > > data split; > > input rep A B yield; > > cards; > > 1 1 1 25.0 > > 1 1 2 31.0 > > 1 2 1 21.5 > > 1 2 2 24.0 > > 2 1 1 41.2 > > 2 1 2 36.1 > > 2 2 1 40.4 > > 2 2 2 37.8 > > 3 1 1 26.3 > > 3 1 2 32.4 > > 3 2 1 28.6 > > 3 2 2 27.6 > > ; > > > proc glm; > > class rep A B; > > model yield= rep A rep*A B A*B; > > test h=A e=rep*A; > > run- Hide quoted text - > > > - Show quoted text - > > This may help > > http://www.ats.ucla.edu/stat/SAS/library/SASExpDes_os.htm Hi data _null_, I already seen that article but it will not solve my problem.Actually what I want to calculate is by using the MSE values of Model Anova table. I have to use these values in the above mentioned formulas to get my result. But how can I calculate it. There is no simple SAS option so that I get these standard errors for split plot. I have to use IML. Please reply.
From: "Data _null_;" on 8 Mar 2010 08:40
On 3/7/10, ManojK <khandelwalmanoj85(a)gmail.com> wrote: > I already seen that article but it will not solve my problem.Actually > what I want to calculate is by using the MSE values of Model Anova > table. I have to use these values in the above mentioned formulas to > get my result. But how can I calculate it. There is no simple SAS > option so that I get these standard errors for split plot. I have to > use IML. You can use ODS to output the MSEs etc and calculate the SEs if you like or you can use PROC MIXED. LSMEANS /DIFF will show the SEs you seek. Or you can write ESTIMATES statements to do the same thing. data split; input rep A B yield; cards; 1 1 1 25.0 1 1 2 31.0 1 2 1 21.5 1 2 2 24.0 2 1 1 41.2 2 1 2 36.1 2 2 1 40.4 2 2 2 37.8 3 1 1 26.3 3 1 2 32.4 3 2 1 28.6 3 2 2 27.6 ;;;; run; proc glm; class rep A B; model yield= rep A rep*a B A*B; random rep*a; test h=a e=rep*a; quit; data _null_; _1 = sqrt(2*8.5633333/(3*2)); _2 = sqrt(2*12.0616667/(3*2)); _3 = sqrt(2*12.0616667/3); _4 = sqrt(2*(1*12.0616667+8.5633333)/(3*2)); put (_all_)(/=); run; proc mixed method=type3; class rep A B; model yield= rep A|B; random rep*a; lsmeans a|b / diff; ** Can we do it with ESTIMATE; *1. Estimate of S.E. of difference b/w two main plot treatment means = Sqrt(2*E1/r*b)=2*8.5633333/3*2=1.6895; estimate 'Sqrt(2*E1/r*b) main plot' a 1 -1 / e; *2. Estimate of S.E. of difference b/w two sub plot treatment means = Sqrt(2*E2/r*a)=(2*12.0616667/3*2=2.00513; estimate 'Sqrt(2*E2/r*a) sub plot' b 1 -1 / e; *3. Estimate of S.E. of difference b/w two sub plot treatment means at the same level of main plot treatment = Sqrt(2*E2/r) = 2.8356; estimate 'Sqrt(2*E2/r) same main plot' b 1 -1 a*b 1 -1 / e; *4. Estimate of S.E. of difference b/w two main plot treatment means at the same or different levels of sub plot treatment = Sqrt((2*[(b-1)*E2 + E1])/ r*b) = sqrt(2(1*12.0616667+8.5633333))/3*2) 2.62202; estimate 'Sqrt((2*[(b-1)*E2 + E1])/ r*b)' a 1 -1 b 1 -1 / e; run; |