Methods? using SPSS 10.0?

Methods? using SPSS 10.0?

Post by shau » Sat, 01 Mar 2003 05:35:37



I am in need of veteran' knowledge I am designing a research
project and I would like to consider the type of analysis/statistics
that I should be using.

Back ground info

A little bit about the project I am a sport psychologist and my
hypothesis is that a well devised training program (use of proper
periodization theory) should increase self-efficacy (confidence):
both: state and trait self-efficacy.

-Furthermore, if the athletes are able to properly "cognitize" this
overall effect that they will have increased self-efficacy
(confidence).

I used three groups:

*All three groups trained 2x per week. The possible confounding
variables were controlled.

1/ The first group is the control group: They were given ONLY the TSCI
and SSCI scales at the beginning of the experiment; at the mid point
of the experiment (three weeks after the start of the experiment) and
at the end of the experiment (three weeks after the 2nd testing).

2/ The second group received a periodized training plan: This group
followed the exact same timing as the first group, in regards to the
administration of the Inventories. The exception  being that they
received a training plan that was based on the Borg RPE scale. Each
training they were told the Borg level that they were to train at. The
exact difference between the two treatments (the first 3 weeks and the
second 3 weeks)  were kept.

3/ The third group received the same exact periodized training plan as
well as kept a training log of their own Borg RPE level which they
reached: Again this group followed the same procedures as the other
two groups in regards to the application of the TSCI and SSCI
inventories. They were also told the Borg RPE level that they were to
train at. However, (the difference being) they kept their own Borg RPE
level which they attained during that training. For, example, they
might have been asked to train at level 17 but only trained at level
15. This is where the cognition comes into play.

-I administered the State and Trait confidence inventories (SSCI and
TSCI) to all three groups prior to the start of the experiment.
-I then gave the SSCI and TSCI at the mid point (after three weeks)
and at the end of the experiment (after three weeks)

Hypothesis

1) That there will be significant differences in confidence (for both
state and trait confidence) between all three groups. Between group
comparison.
2) There will be a significant difference in confidence levels  (for
both state and trait confidence) within the training group and the
borg group.
3) That there will be significant differences in confidence between
gender within each group.
4) That the 3rd group (the Borg group) will differ in significance
levels of confidence dependant on whether or not they achieved their
training goals.
5) That the questions used to define self-efficacy' will support the
theory of self-efficacy.

Design

I use three measurements for the experiment:

*There is only one level for each of the scales. They both test only
trait (TSCI) or state (SSCI) confidence in sport.
1) Trait Sport Confidence Inventory (TSCI): This is a 13 question
inventory which uses a interval (Likert) scale that goes from 1-9.
2) State Sport Confidence Inventory (SSCI): That is a 13 question
inventory which uses a interval (Likert) scale that goes from 1-9.
3) Borg Rating of Perceived Exertion Scale (Borg RPE): This is
interval scale that goes from 6 to 20. It is a one shot scale that is
used directly after exercise. The scale is actually rated from 1-20
but the first 5 measures are used for levels less than actual movement
(for example, trance or sleep)

And I used 7 questions to further understand the relationship between
state and trait confidence and the use of Borg's scale for goal
setting.

1) Perceived success: A percentile score out of 100.
2) Years of training: Years
3) Actual success: Results scores: 3 points =1st place, 2 points = 2nd
place and 1 point = 3rd place.
4) # of tournaments: numerical
5) Competitions in and out of dojo: dummy coded. 1 = yes and 2 = no
6) Difference in set Borg levels and actual Borg level attained in
first training block (goal setting difference). Dummy coded: 1 = no
change (all goals met), 2 = negative change (did not reach goals) and
3 = positive change (exceeded goals set).
7) Difference in set Borg levels and actual Borg level attained in
second training block (goal setting difference). Dummy coded: 1 = no
change (all goals met), 2 = negative change (did not reach goals) and
3 = positive change (exceeded goals set).

Statistics Packages Available / Used

-SPSS 10.
-Student version EQS (I will be buying the full version soon)
-Student version of LISREL and HLM
-WinBugs (HLM)

Questions

So now I will place the questions

First Hypothesis. That there will be significant differences in
confidence (for both state and trait confidence) between all three
groups. Between group comparison.

-To test the main effect between the 3 groups, I used a Repeated
measures ANOVA.

Q: Is there are better way to analysis the variance between the
groups?

Second Hypothesis. There will be a significant difference in
confidence levels  (for both state and trait confidence) within the
training group and the borg group.

-Again I used a repeated measures ANOVA.

Q: and again would another stat be better suited to judge differences.
I am wondering if I shouldn't use ANCOVA with some of the questions I
used to test for relationships to confidence. I have strong
theoretical knowledge in this area and wouldn't be taking a shot in
the dark as far as choosing the covariate.

Third Hypothesis. That there will be significant differences in
confidence between gender within each group.

-* I have many questions as to which method to use. I have split the
file and used Repeated measures ANOVA. However, I am also thinking
about either Discriminant Analysis or Logistic regression.

Q: Please advise as to what is the most useful statistic in regards to
the question above?

Fourth Hypothesis. That the 3rd group (the Borg group) will differ in
significance levels of confidence dependant on whether or not they
achieved their training goals.

-This is easily the most complex area of my analysis (for me). As I
have two dependant variables (the State confidence level or the Trait
confidence level and the Borg level. ***Note that the difference
between both levels of confidence are not important to me), I decided
to use MANOVA and then discriminant analysis.
-However, when I looked at the data I could categorize the Borg stats
as to whether or not they achieved the training goals set of not. I
could then Dummy code the sample into those who have exceeded their
goals and those that did not reach their goals  (there was a code for
those who met their goals exactly, however no one was tabulated for
this code).

Q: After reviewing a good amount of the conversations/literature
regarding the possible research statistics to use I am in a utter
stat of confusion as there are some many methods to use which all
could answer my questions on this hypothesis.

Fifth Hypothesis. That the questions used to define self-efficacy'
will support the theory of self-efficacy.

-I started out with multiple regression analysis as I have a fairly
good amount of theoretical background to be able to decide which
method to use for inputting the predictor variables.

Q: I am currently learning about SEM and have the student versions of
the major packages. Would a SEM model be more useful than a regression
model?
I thank everybody who decides to help me with these answers in
advance. Thank-you for your time and consideration. (When I come to
sites like this, it truly amazes me how helpful people are).

Sincerely, Shaun Galloway