Hi Yuanchao -<div><br><div>Sorry for the delay.<br>
<br><br><div class="gmail_quote">On 27 April 2010 04:10, Zhangyuanchao <span dir="ltr">&lt;<a href="mailto:[log in to unmask]"><a href="/cgi-bin/wa-jisc.exe?LOGON=A3%3Dind1004%26L%3DFSL%26E%3Dbase64%26P%3D2342924%26B%3D--001485f7c4b0f69048048571270b%26T%3Dtext%252Fhtml%3B%2520charset%3DISO-8859-1%26pending%3D" target="_parent" >[log in to unmask]</a></a>&gt;</span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex;">

<table cellspacing="0" cellpadding="0" border="0"><tbody><tr><td valign="top" style="font:inherit"><div><div></div><div class="h5"><div>Dear FSL experts,</div>
<div> </div>
<div>      I want to make three group (disease1, disease2 and normal control) comparisions with two covariants (age and gender), that is, ANCOVA. However, I am not sure the design matrix and the contrasts are accurate. Could you please hep me examine them? For example, I have 10 samples and I want to answer whether there are significant differences between any of two groups and then post-hoc t test which two groups the differences exist in. I set the design matrix and contrasts as follows:</div>


<div> </div>
<div>Design Matrix:</div>
<div> </div>
<div>Group  EV1(D1)  EV2(D2) EV3(NC) EV4(Age) EV5(Gender)</div>
<div>1           1               0           0          22            1</div>
<div>1           1               0           0          25            0</div>
<div>1           1               0           0          23            1</div>
<div>1           0               1           0          22            1</div>
<div>1           0               1           0          26            0</div>
<div>1           0               1           0          24            0</div>
<div>1           0               0           1          23            1</div>
<div>1           0               0           1          24            0</div>
<div>1           0               0           1          26            1</div>
<div>1           0               0           1          23            1</div>
<div> </div>
<div>Design Contrasts:</div>
<div>Total differecese:</div>
</div></div><div>1        1          1          0      0        F-test</div>
<div>Other t-tests:</div>
<div>1        -1         0          0      0</div>
<div>-1        1         0          0      0</div>
<div>0         1        -1          0      0</div>
<div>0         -1        1          0      0</div>
<div>1         0        -1          0      0</div>
<div>-1        0         1          0      0</div>
<div> </div></td></tr></tbody></table></blockquote><div><br></div><div>Close.  You pretty much want to follow the ANOVA 1-factor 4-levels example in the FEAT details examples (obviously removing the extra level), and then add columns for the age and gender covariates (you should demean these).  </div>

<div><br></div><div>Depending on your hypotheses and how the diseases are related, another option could be simply to first have a regressor modelling difference between disease and non/disease and then a regressor modelling the difference between the two diseases, and doing separate t/f tests on these.</div>

<div><br></div><div>Cheers,</div><div><br></div><div>Eugene</div><div><br></div><div><br></div><div><br></div><div> </div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex;">

<table cellspacing="0" cellpadding="0" border="0"><tbody><tr><td valign="top" style="font:inherit">
<div>Do you think it is correct for the design matrix and contrasts?</div>
<div> </div>
<div>Thank you very much for your help!</div>
<div> </div>
<div>Best,</div>
<div> </div>
<div>Yuanchao</div>
<div> </div></td></tr></tbody></table><br>






       </blockquote></div><br></div></div>