MATLAB Answers

repeated measure anova two options

25 views (last 30 days)
Sophi Af
Sophi Af on 15 Dec 2020
Commented: Sophi Af on 24 Dec 2020
Hello,
I am trying to use repeated measure ANOVA “ranova” in MATLAB. According to MATLAB tutorial, there are two option: 1: ranova(rm) and 2: ranova(rm,'WithinModel',WM).
I tried both and they give different results for the within model analysis (in my example the within model variable is time; I have repeated measurements). Option 1 has a closer results to those produced by other software like R; however, it does not give me the between model results which are the main effect of between model variables and their interactions. I appreciate it if you let me know what the difference is.
Also, the sums of squares are not the same as what I get from R. Could you please help me find out about this discrepancy?
Best,
Sepideh.

  1 Comment

Sophi Af
Sophi Af on 15 Dec 2020
I have included my code here. The data is measurement of mice brain at 8 regions; thus, I am running 8 ranova.
tr=cell(length(tr_p),1);
gr=cell(length(tr_p),1);
for i=1:length(tr_p)
tr{i}=int2str(tr_p(i));
gr{i}=int2str(gr_p(i));
end
for i=1:8
Y1=DaB_a1(:,i);
Y2=DaB_a2(:,i);
Y=[Y1,Y2];
tr_cat=categorical(tr);
gr_cat=categorical(gr);
varnames = {'treatment','disorder','A1','A2'};
t=table(tr_cat,gr_cat,Y1,Y2,'VariableNames',varnames);
% Time=[1 2];
T = table([1 2]','VariableNames',{'Time'});
rm=fitrm(t,'A1-A2 ~ treatment*disorder','WithinDesign', T);
% rm=fitrm(t,'A1-A2 ~ treatment*disorder','WithinModel', Time, 'WithinModel', 'separatemeans');
% [ran0] = ranova(rm);
[ran1] = ranova(rm,'WithinModel','Time');
[ran0] = ranova(rm);
adj1=epsilon(rm);
M1=multcompare(rm, 'Time', 'By', 'treatment');
M2=multcompare(rm, 'Time', 'By', 'disorder');
tr_cat=categorical(tr);
gr_cat=categorical(gr);
intr_cat=tr_cat.*gr_cat;
varnames_cat = {'intr_cat','A1','A2'};
t=table(intr_cat,Y1,Y2,'VariableNames',varnames_cat);
rm=fitrm(t,'A1-A2 ~ intr_cat','WithinModel', Time, 'WithinModel', 'separatemeans');
% [ran2] = ranova(rm);
[ran3] = ranova(rm,'WithinModel','Time');
adj3=epsilon(rm);
M3=multcompare(rm, 'Time', 'By', 'intr_cat');
1;
end

Sign in to comment.

Answers (1)

Jeff Miller
Jeff Miller on 15 Dec 2020
Not sure but I think you need
T = table([1 2]','VariableNames',{'Time'});
T.Time = categorical(T.Time); % add this line
Maybe post one of your 't' data tables so the design is clearer?

  3 Comments

Sophi Af
Sophi Af on 23 Dec 2020
Hello,
Thank you for your reply. I guess my main issue is the difference between the results of two option: 1: ranova(rm) and 2: ranova(rm,'WithinModel',WM).
I have attached one of my t table data to this email. The data has been recorded in 8 brain areas but I am looking into only the first area now (first data sheet).
I appreciate your help.
Best,
Sepideh.
Jeff Miller
Jeff Miller on 24 Dec 2020
If each line of your t table represents a different person, then you should use the ranova command including 'WithinModel'.
Sophi Af
Sophi Af on 24 Dec 2020
Unfortunately, when I use the command including 'WithinModel' the results doesnt make sense including the degrees of freedom. Furthermore, the results are going to be far from what I get from other softwares like R.

Sign in to comment.

Tags

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by