<P>step 1: logrithms daily DAX-index, LNDAX=log(dax)</P>
<P>step 2: unit root test, select level, ADF is greater than crical value, H0 is rejected</P>
<P>step 3: unit root test, select first order difference, ADF is less then critical value at 99% confidence, H0 can not be rejected (-51.03&lt;-3.96), then non unit root, it is now stationary.</P>
<P><FONT color=#3809f7><b>step 4: i create a new serie: E=LNDAX-mean(LNDAX</b></FONT>), i am not sure if it is right , or shall i differece E at first order???</P>
<P>step 5: command ls e ar(1) ma(1)</P>
<P>step 6: result with R</P>
<P>attatched document is my data.</P>

<P>For another question, the result with ARMA(1,1) is not far better than OLS with LNDAX=c+c(1)+LNDAX(-1 to -3), they are all over 99.8%, is ARMA process here necessary?</P>
<P>Thanks in advance!!!</P>