Class Notes 28 Oct 2004
tips for Results and Analysis section of
writeups
EPSY
6301
28 Oct 2004
Results and Analysis Section tips
- must be a graduate level writeup
you seldom see symbols in an article
Principle 1: Organize around research hypotheses stated in the intro, describe results and analysis for 1st hypothesis, then 2nd, etc.
Principle 2: don't need to show formulas, just name standard statistical procedures
Principle 3: Only stats based on raw scores are usually reported, not raw scores
Principle 4: Present descriptive stats first, usually start with central tendency and variablity
Principle 5: Organize large amounts of data in tables, give each table a number and descriptive title
- table does not stand alone, you still need to explain what is in the table
- add some narrative
- boldface the salient features (most important parts)
look at APA style tables
Principle 6: describe main conclusions based on each table, point out highlights that reader may overlook, do not discuss each table entry
can move control from the end to the front, then that can make your graph gradually go up (ascending order)
- this is one of the best things about the computer, ability to move / change variable order
Principle 7: Stat figures should be professionally drawn and used sparingly.
need tables:
- statistical
- inferential
- factor analysis data
Principle 8: stat sybmols should be underlined or italicized
Principle 9: Use proper case for each stat symbol
Principle 10: Numerals that start sentences and are less than 10 should be spelled out
Principle 11: Qualitative results need to be organized and the organization made clear to the reader
RESULTS SECTION IS JUST THE FACTS / THE DATA
Discussion session includes:
- brief summary of the study
- analysis of the results
Principle 1: include brief summary, in dissertation include formal summary
Principle 2: Restate hypotheses from intro and indicate if they were supported by the data
Principle 3: Highlights of results should be described
- highlight #1, #2, etc.
- in our first project, we just have 1 highlight
- in 2nd project, also want an analysis of the blocking, did we control for the nuisance variable, etc.
Principle 4: explicitly state implications
- if you don't have lit review backing you up, it is harder to discuss implications
- for us now in these projects, we can BS some....
Principle 5: identify strengths and limitations
- include limitations / weaknesses too
Principle 6 (not applicable for us): Issue a call for future research, provide specific guidance
Principle 7: Point out consistencies and inconsistencies of results with current literature
- this is difficult for us now
Principle 8: Speculate
- this is the only place you can speculate
- this is going beyond what the data could have told us (be careful and don't push it, stay within the parameters of the data)
Principle 9: Inappropriate to intro new data/references in discussion and conclusion section
For most of us (14 of the class) we need to redo our discussion/analysis for project 1
- 6 students got an A already on project 1
- don't give project 2 tonight: we should fix it first
- turn in everything
- at end of semester, we should have a portfolio of work we have done in this class
-- turn in the excel and SPSS
make sure you include PRACTICAL significance discussion
This will be our last night on Kirk, after tonight we will start Tabashnik
- each chapter will be highlighed
2 projects will be coming from Tabashnik
- multiple regression
- MANOVA
We WILL MEET the last night (scheduled as a final) where students can present their project results (about 10 min each)
- This should be a 25 page comprehensive paper that includes all components just like a dissertation
- Intro
- Summary of Lit Review
- Methodology
- Results
- Discussion and Conclusions
- References and Appendices
certain key tables WILL go in the body
- auxillary tables will go in the appendices
APA style, 1 inch margins, double margins, 12 point, etc.
This is like your first draft of an article submission
- Dr O is working on a rubric for what needs to be there (general rubric)
We want to rush a little bit now so the last few meetings we don't have additional projects, but are having lectures from Tabachnik
- next 3 meetings will focus on chapter projects
- last 3 meetings will be a show and tell while we are working on our own projects
WE NEED TO START TURNING IN SOME ARTICLE CRITIQUES
- these should be linked to our project
- kill 2 birds with 1 stone by using these as part of our lit review
this class is a real PhD level class, a lot of work!
- this class has more people (about) from outside COE than within
3 WAY ANALYSIS OF VARIANCE
- if you are doing this for your dissertation, then this is more important, otherwise don't worry as much of this
How do you do test for additivity in SPSS
-
include this data in our project writeup
Look at p 270, look at data
- enter column 1 as subject number
- enter treatment groups in subsequent columns
- enter variable view to add labels
To get non-additivity:
- Analyze - Scale - Reliability Analysis
-- enter treatments as items
-- click on statistics: ask for descriptives you want, then click on TUKEY'S TEST FOR ADDITIVITY
- click to get intraclass correlation, get consistency
Tukey test for multiple comparisons is different than Tukey's test for additivity
- results for this will be labeled under ANOVA with Friedman's Test and Tuckey's Test for Nonadditivity
What does the sum of squares of non-additivity
- this tells us if there is an interaction effect
- we want to show it is non-additive
usually for additivity you want to use an alpha of .1 because you want to make it easy to reject, hoping that you won't
- if you don't reject, then the F for non-additivity is not significant so we conclude the interaction effect is negligible (assume it is zero)_
-- then you can avoid in your model having a multiplicative term
- once you rule it out, you throw it out
you want to rule this out because it makes the formula simpler, and therefore the effects/results you find less complicated to explain
In SASS you have to program it to do this, you can't do it as easily as you can in SPSS
Dr O still has yet to find more than 10 errors in the Kirk book
- that is remarkable, because most statistics books have at least 100 errors
Kirk has never gone beyond between subjects measures, is missing some things like repeated measures
In 2 way ANOVA, we have a hypothesis about the interaction between treatment levels, so we do not want to rule this out / get rid of it as above
- the interaction is a hypothesis that needs to be tested, along with main effects
- there are 3 main hypotheses
- if the interaction is significant, this compounds the complexity of the results and interpretation
- one way to interpret the significance of the intraction is to look at the plots (let SPSS do it or just plot cell means)
- 2nd alternative: do simple main effects (look at each cell level and see if we can create an F to determine where the interaction is coming from)
plot can tell us where the interaction is coming fom
- can be misled sometimes by SPSS or other charts if the origin is not set to zero
- not good to rely only on the plots
- also good to use statistical tests
p. 377 start to talk about single main effects
- further partitioning main effects as they relate to the interaction
-
p. 382 talks about the 2nd approach (SEP)
How do we do this in SPSS?
28 Oct 2004
Results and Analysis Section tips
- must be a graduate level writeup
you seldom see symbols in an article
Principle 1: Organize around research hypotheses stated in the intro, describe results and analysis for 1st hypothesis, then 2nd, etc.
Principle 2: don't need to show formulas, just name standard statistical procedures
Principle 3: Only stats based on raw scores are usually reported, not raw scores
Principle 4: Present descriptive stats first, usually start with central tendency and variablity
Principle 5: Organize large amounts of data in tables, give each table a number and descriptive title
- table does not stand alone, you still need to explain what is in the table
- add some narrative
- boldface the salient features (most important parts)
look at APA style tables
Principle 6: describe main conclusions based on each table, point out highlights that reader may overlook, do not discuss each table entry
can move control from the end to the front, then that can make your graph gradually go up (ascending order)
- this is one of the best things about the computer, ability to move / change variable order
Principle 7: Stat figures should be professionally drawn and used sparingly.
need tables:
- statistical
- inferential
- factor analysis data
Principle 8: stat sybmols should be underlined or italicized
Principle 9: Use proper case for each stat symbol
Principle 10: Numerals that start sentences and are less than 10 should be spelled out
Principle 11: Qualitative results need to be organized and the organization made clear to the reader
RESULTS SECTION IS JUST THE FACTS / THE DATA
Discussion session includes:
- brief summary of the study
- analysis of the results
Principle 1: include brief summary, in dissertation include formal summary
Principle 2: Restate hypotheses from intro and indicate if they were supported by the data
Principle 3: Highlights of results should be described
- highlight #1, #2, etc.
- in our first project, we just have 1 highlight
- in 2nd project, also want an analysis of the blocking, did we control for the nuisance variable, etc.
Principle 4: explicitly state implications
- if you don't have lit review backing you up, it is harder to discuss implications
- for us now in these projects, we can BS some....
Principle 5: identify strengths and limitations
- include limitations / weaknesses too
Principle 6 (not applicable for us): Issue a call for future research, provide specific guidance
Principle 7: Point out consistencies and inconsistencies of results with current literature
- this is difficult for us now
Principle 8: Speculate
- this is the only place you can speculate
- this is going beyond what the data could have told us (be careful and don't push it, stay within the parameters of the data)
Principle 9: Inappropriate to intro new data/references in discussion and conclusion section
For most of us (14 of the class) we need to redo our discussion/analysis for project 1
- 6 students got an A already on project 1
- don't give project 2 tonight: we should fix it first
- turn in everything
- at end of semester, we should have a portfolio of work we have done in this class
-- turn in the excel and SPSS
make sure you include PRACTICAL significance discussion
This will be our last night on Kirk, after tonight we will start Tabashnik
- each chapter will be highlighed
2 projects will be coming from Tabashnik
- multiple regression
- MANOVA
We WILL MEET the last night (scheduled as a final) where students can present their project results (about 10 min each)
- This should be a 25 page comprehensive paper that includes all components just like a dissertation
- Intro
- Summary of Lit Review
- Methodology
- Results
- Discussion and Conclusions
- References and Appendices
certain key tables WILL go in the body
- auxillary tables will go in the appendices
APA style, 1 inch margins, double margins, 12 point, etc.
This is like your first draft of an article submission
- Dr O is working on a rubric for what needs to be there (general rubric)
We want to rush a little bit now so the last few meetings we don't have additional projects, but are having lectures from Tabachnik
- next 3 meetings will focus on chapter projects
- last 3 meetings will be a show and tell while we are working on our own projects
WE NEED TO START TURNING IN SOME ARTICLE CRITIQUES
- these should be linked to our project
- kill 2 birds with 1 stone by using these as part of our lit review
this class is a real PhD level class, a lot of work!
- this class has more people (about) from outside COE than within
3 WAY ANALYSIS OF VARIANCE
- if you are doing this for your dissertation, then this is more important, otherwise don't worry as much of this
How do you do test for additivity in SPSS
-
include this data in our project writeup
Look at p 270, look at data
- enter column 1 as subject number
- enter treatment groups in subsequent columns
- enter variable view to add labels
To get non-additivity:
- Analyze - Scale - Reliability Analysis
-- enter treatments as items
-- click on statistics: ask for descriptives you want, then click on TUKEY'S TEST FOR ADDITIVITY
- click to get intraclass correlation, get consistency
Tukey test for multiple comparisons is different than Tukey's test for additivity
- results for this will be labeled under ANOVA with Friedman's Test and Tuckey's Test for Nonadditivity
What does the sum of squares of non-additivity
- this tells us if there is an interaction effect
- we want to show it is non-additive
usually for additivity you want to use an alpha of .1 because you want to make it easy to reject, hoping that you won't
- if you don't reject, then the F for non-additivity is not significant so we conclude the interaction effect is negligible (assume it is zero)_
-- then you can avoid in your model having a multiplicative term
- once you rule it out, you throw it out
you want to rule this out because it makes the formula simpler, and therefore the effects/results you find less complicated to explain
In SASS you have to program it to do this, you can't do it as easily as you can in SPSS
Dr O still has yet to find more than 10 errors in the Kirk book
- that is remarkable, because most statistics books have at least 100 errors
Kirk has never gone beyond between subjects measures, is missing some things like repeated measures
In 2 way ANOVA, we have a hypothesis about the interaction between treatment levels, so we do not want to rule this out / get rid of it as above
- the interaction is a hypothesis that needs to be tested, along with main effects
- there are 3 main hypotheses
- if the interaction is significant, this compounds the complexity of the results and interpretation
- one way to interpret the significance of the intraction is to look at the plots (let SPSS do it or just plot cell means)
- 2nd alternative: do simple main effects (look at each cell level and see if we can create an F to determine where the interaction is coming from)
plot can tell us where the interaction is coming fom
- can be misled sometimes by SPSS or other charts if the origin is not set to zero
- not good to rely only on the plots
- also good to use statistical tests
p. 377 start to talk about single main effects
- further partitioning main effects as they relate to the interaction
-
p. 382 talks about the 2nd approach (SEP)
How do we do this in SPSS?
Posted: Mon - November 8, 2004 at 08:24 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: 8
Published On: Nov 08, 2004 08:25 PM
Total entries in this category: 8
Published On: Nov 08, 2004 08:25 PM
