You have N=156 and N=157 for "Phase 1" and Phase 3", so it
*looks* like you have done this:
You did a factoring on the combined set, and let SPSS score the
factors. This incorporates every item into every factor, with a
smaller or greater weight. The direction of the scale / factor
is arbitrary technically -- observe the direction of the largest
loadings to figure what direction the meaning is.
(For useful scales, the usual practice in clinical work is
to use a scoring rule: make a scale by creating the average
item score for only the items that load above some cutoff, with
the cutoff set at 0.35 or 0.40 or 0.45 or whatever, as needed in
order to obtain a single factor for each variable to "belong to."
Starting with "too many variables" you will get relatively large
loadings.)
What you have presented seems to be the means, etc., for the
two phases that you started out with. As David mentioned, the
overall mean for factor scored by the FA is 0.0, with variance of 1.0;
that seems to be what your table shows, for data divided into two
Phases. The differences between phases range from about 0.6,
which is fairly large, to less than 0.2, which is fairly small (though
a paired t-test will probably still show Phases as different).
Since this is paired data, you would want a paired t-test to
compare the two Phases properly.
Possibly a majority of statisticians in the world have never done
a factor analysis of any kind. They do teach it in departments of
Education and Psychology. I hope it might be part of Biostatistics
in Schools of Public Health these days, but it was not on the
curriculum when I learned my formal statistics. I don't remember
economists ever mentioning FA. What with "data mining", it
ought to be taught more these days that it used to be, but I
don't know whether it is. - This is in preparation to my saying,
"You need a statistician who knows about this sort of statistics."
I learned most of my data analysis by reading some excellent
articles and a couple of textbooks while working on a job where
I had an expert mentor on statistics; and then another job with
expert mentor on experimental design. After a few years, I went
back to school and picked up a background in formal theory, which
was more useful than I had imagined it would be.
In short: You don't know what you are doing? You won't create
a decent job of it from throwing a stat-pack at the data unless
you have pretty good feedback from a decent statistician who
knows about the tools that you need.
--
Rich Ulrich
[table format revised, below]
> Date: Tue, 23 Oct 2012 19:59:12 -0700
> From:
[hidden email]> Subject: Deciphering from t test the mean
> To:
[hidden email]>
> Hi,
> Can someone help me as to why the mean is negative? How does the negative
> come about? Someone said that this would essentially indicate that there are
> no significant group differences with respect to these variables based on
> the data analyzed. I fed this into SPSS but what is the mathematics behind
> it? I mean how is this mean calculated when SPSS does this? I thought it is
> the total scores of say 1-SD, 2-D, 3- N, 4-A, 5-SA and divide that with the
> number of factors that I have with e.g. Conf Level Factor. Shouldnt this be
> positive?
>
> Table 1: Factor Scores: Group Statistics
> Measure Phase N Mean Std. Dev. S.E.M.
Conf. Level Factor 1.00 156 -0.199 0.934 0.075
3.00 157 0.185 1.020 0.081
> Feedback Assess.
Factor 1 1.00 156 -0.219 0.977 0.078
> 3.00 157 0.208 0.976 0.078
> Feedback Assess.
Factor 2 1.00 156 -0.091 0.981 0.079
> 3.00 157 0.095 1.015 0.081
>
> Thanks
...
Locked
