"Pingu" <[EMAIL PROTECTED]> schreef in bericht
news:[EMAIL PROTECTED]
> Ok.....sort of...
>
> Im still not fully clear, but rest assured you have already done a
far
> superior job to my tutors.
>
> Lets say that VD dialling and TST dialling in quiet conditions both
> show a very similar performance in terms of dialling speed. When
noise
> is introduced, however, VD becomes much slower than TST because the
> voice recogniser has a lot of interference to deal with.
>
> My 2 t-tests would show a non-significant difference between the
> systems in quiet conditions and a significant difference between
them
> in noisy conditions.
>
> What will the ANOVA show? So far i just cant see how the results
from
> an ANOVA can describe such a clear distinction as this.

The analysis of your data should start with 12x4 tables of means per
subject for each of your 4 treatment combinations (dialing methods x
noise conditions). It is perfectly valid to compare any two treatment
combinations using a t-test for paired data, or equivalently, using a
t-test on the differences, calculated per subject. What ANOVA (with 3
factors: dialing methods, noise level and subject as a blocking
factor) may add is a more efficient estimate of error variance, if
this variance is believed to be the same for all treatments (which
appears to be doubtfull in your case).

>
> p.s - It would really help me if you framed your anwer in terms of
the
> output i will get from SPSS. When i ran the ANOVA i did 2 factors
> (noise/quiet) with two levels (TST/VD). The output gives three
> figures:
>
> Noise = X
> Quiet = X
> Noise*Quiet = X
>
This is nonsense. Your factors are dialing methods and noise level,
each with two levels (plus the above mentioned blocking factor
subjects!).

.
.
=================================================================
Instructions for joining and leaving this list, remarks about the
problem of INAPPROPRIATE MESSAGES, and archives are available at:
.                  http://jse.stat.ncsu.edu/                    .
=================================================================

Reply via email to