You can't draw this graph especially easily.
The confidence interval that is relevant for a t-test is the
confidence interval of the difference between the two means, not the
confidence interval of each mean.
However, for a t-test, I would wonder whether a graph is useful. You
are only trying to show two numbers. You can show those two numbers
in text, or in a table, and you can give the difference between them
and the confidence interval, which is what you're interested in.
I would write something like: Group 1 had a mean value of 17 (SD 4),
group 2 had a mean of 19 (SD 3), thus group 2 had a mean which was 2
points higher (95% CIs 1, 3; p = 0.002).
Jeremy
On 21/04/2008, Mann, Helen N <[log in to unmask]> wrote:
>
>
>
> If you had done bar charts to show differences in mean scores for males and
> females on 3 measures and you had indicated on your chart with a star where
> the t-test had found significant sex differences (p<.001 and p<.05) but then
> your supervisor said you need to show confidence intervals on your graphs
> what would you do?!!?
>
> In the output from T-tests a 95% confidence interval is calculated...so
> would you report that, if so how would you report it?!.....But surely,
> shouldn't the person looking at the graph just assume that as you did a
> t-test a 95% confidence interval level was taken into account so you don't
> need to report it....?!?!
>
> Or would you scrap the bar chart altogether and just do an error bar graph?!
> Or show both....
>
> Any help would be most appreciated...(hope it all makes sense!)
>
> H*
> (PS>> If there are any SPSS stats gurus in Edinburgh please contact me!)
> ________________________________
> Heriot-Watt University is a Scottish charity registered under charity
> number SC000278.
>
--
Jeremy Miles
Learning statistics blog: www.jeremymiles.co.uk/learningstats
Psychology Research Methods Wiki: www.researchmethodsinpsychology.com
|