Notes 2 Dec 2004
Overview of MANOVA
EPSY
6306
12-2-2004
If you are doing a MANOVA procedure or multiple regression, you don't need to do project #5
- not a big deal: just run the problem and see what you know and learned about the procedure
MANOVA example: 3 different dependent variables for each case, not repeated measures
He will work a factorial MANOVA example
Friday Dec 17th is when the final project is DUE!!!!!!!
No meeting on Dec 16th, instead we are meeting next week on dead day (Dec 9th)
In MANOVA we are acknowledging that there are 3 possible outcome variables
- reasonable numbers of levels for IV: 3-5
- if you have 5 levels, you might as well do a multiple regression
don't like to use age as an IV because it really isn't a "treatment"
always more powerful to stick with continous data rather than go to interval level (categorized) data
the temporal aspect of causality is very important
- the cause must take place before the effect
orthagonality = no correlation
reason we do MANOVA is because the variables we are looking at are somewhat correlated
- MANOVA is based on experimental design not
covariate idea is to equalize the groups even when there is random assignment
- sometimes it is used when people cannot use random assignment
let the literature guide you whether you use a variable as a covariate or an IV
- if you forgot to look or include a particular variable in your study, then you can use that as a covariate
-- years of experience, age, courses taken: all are good covariate (must be a continuous value)
MANOVA is efficient: does 2 ANOVAs at once and also takes into account how the variables are interrelated
- not bivariate: multiple variables that are somehow related to each other
- example: number of drownings at a beach and soda sales: they may be related at .95, but if you want to go to a predictive study (also temperature, numbers of people at the beach)
in Education .4 can be a large significance, may never see .7
Tukey is really multiple t tests
main effects are better explained through interactions rather than main effects generally
MANOVA SPSS results
- don't worry about the intercept line (we ignore that)
Can have good studies with 12 or 15 cases per level
- generally you want between 10 and 20 cases per level
- if you have random assignment and random selection, you can go as low as 5 in an experiment
See checklist for MANOVA on page 383 in Tabaschnik
12-2-2004
If you are doing a MANOVA procedure or multiple regression, you don't need to do project #5
- not a big deal: just run the problem and see what you know and learned about the procedure
MANOVA example: 3 different dependent variables for each case, not repeated measures
He will work a factorial MANOVA example
Friday Dec 17th is when the final project is DUE!!!!!!!
No meeting on Dec 16th, instead we are meeting next week on dead day (Dec 9th)
In MANOVA we are acknowledging that there are 3 possible outcome variables
- reasonable numbers of levels for IV: 3-5
- if you have 5 levels, you might as well do a multiple regression
don't like to use age as an IV because it really isn't a "treatment"
always more powerful to stick with continous data rather than go to interval level (categorized) data
the temporal aspect of causality is very important
- the cause must take place before the effect
orthagonality = no correlation
reason we do MANOVA is because the variables we are looking at are somewhat correlated
- MANOVA is based on experimental design not
covariate idea is to equalize the groups even when there is random assignment
- sometimes it is used when people cannot use random assignment
let the literature guide you whether you use a variable as a covariate or an IV
- if you forgot to look or include a particular variable in your study, then you can use that as a covariate
-- years of experience, age, courses taken: all are good covariate (must be a continuous value)
MANOVA is efficient: does 2 ANOVAs at once and also takes into account how the variables are interrelated
- not bivariate: multiple variables that are somehow related to each other
- example: number of drownings at a beach and soda sales: they may be related at .95, but if you want to go to a predictive study (also temperature, numbers of people at the beach)
in Education .4 can be a large significance, may never see .7
Tukey is really multiple t tests
main effects are better explained through interactions rather than main effects generally
MANOVA SPSS results
- don't worry about the intercept line (we ignore that)
Can have good studies with 12 or 15 cases per level
- generally you want between 10 and 20 cases per level
- if you have random assignment and random selection, you can go as low as 5 in an experiment
See checklist for MANOVA on page 383 in Tabaschnik
Posted: Thu - December 2, 2004 at 09:45 PM
Quick Links
Calendar
Sun | Mon | Tue | Wed | Thu | Fri | Sat |
Categories
Archives
XML/RSS Feed
Statistics
Total entries in this blog:
Total entries in this category: 11
Published On: Dec 02, 2004 10:05 PM
Total entries in this category: 11
Published On: Dec 02, 2004 10:05 PM
