Standard error from glm

[Miracle]

Hi everyone.

I'm wondering if anyone could shed some light on this.

I don't understand why the confidence interval of the least squares means I manually calculated as

(logconc LSMEAN) plus/minus (1.96*Standard Error) is different from that produced by SAS.

My code is as below. Your help is greatly appreciated!

proc glm data=out order=internal;

  by Intervention;

  class Sample;

  model logconc=Sample / solution;

  lsmeans Sample / cl pdiff tdiff stderr e om;

run;

quit;

[PaigeMiller]

There's no way we can answer your question.

We can't see your data, we can't see the calculations you did, and we can't see the SAS output.

[lvm]

Paige is right. One obvious thing: "1.96" is the t value with a large number of denominator df (say, df >30). The critical t value will be larger for small df.

[Miracle]

Thanks very much Paige and Ivm for your reply

Yes, I did it wrong by multiplying by 1.96 :smileysilly: