Class notes 9-2-2004
Overview of the class, review of the syllabus,
tips for article critiques
EPSY
6301
2 September 2004
Get the new Green SPSS book and return the yellow one
We don't do experimental design research in education, but many other fields do
about half our course will be on multivariate analysis
- univariate and multivariate analysis is our focus all term
will use some of the exercises in the Kirk book, because our author (Tabachnick) does not have exercises
- Kirk book is available in the LRC to use but not check out
some of these problems here can take 20 pages to hand write out
There are no tests here, just projects that you work on at home (see last page of syllabus)
- you will also do the same on the computer, and explain how the computer results match the hand calculations
we will NOT do written report (like about 12 pages) instead we will have a 1 page impact / implications page
recommendation is to use computer programs and then work it by hand, so you always have the answer
Arturo likes SASS because he likes to write syntax
Graduate pack for SPSS will help us get through the first 4 projects, but probably won't do multivariate
Typical flaws in a study
1- sample size
2- selection of subjects (simple random sampling, etc --> something that gives you a probability sample) - convenience sampling is always a flaw
- power analysis: is it representative / unbiased group? (must have for good statistical estimates, predications, interpretations, and conclusions)
3- instrument: how do I know the instrument is valid, produces valid scores?
-- is it public domain or commercial? (most commercial instruments are good)
--- many public domain instruments are bad
--- generally have a lot of psychometric studies done on commercial instruments
reliability and validity are the psyometric properties that are key
- validity evidence is often more rare
-- reliable is a close group of shots on a target
-- valid result is the group is dead center on the bullseye
when you read the article, you will go back and really focus on the methodology section
- sins of omission and sins of comission
- if you don't find many problems, then you can put a lot of trust in the results
RESEARCH DESIGN is all about how I control for extraneous variation
in quantitative research, we are trying to find/make causal comparisons
- looking for the causes and consequences (that is our goal)
when you find a problem: explain the consequences
ABESCO tends to do real well with fulltext article versions
look for data driven articles
- not just a review
- need to have methodology, results, etc.
We will do our article reviews from journals in our area of interest / study
FOCUS ON THE CONSEQUENCE: WHY YOU THINK IT IS GOOD OR BAD, IN TERMS OF THE METHOD'S EFFECT ON RESULTS
When you do a search on your topic of your dissertation, if you get 3000 results then you may need something else (that subject may have been thoroughly researched)
Don't fall into the trap in your chapter 2 for moving old bones from a cemetary into your dissertation
- it is not about bulk
- tough to be a 70 page wonder
this course is also about writing research pieces
- we are focused on results and interpretation
For every project we do, we have access to the book, the computer, and Dr O (he has the answer for every project we are working on)
- you can always touch base with him and see if you are doing the calculations correctly so far
Get very familiar with one statistical tool: ANCOVA, MANCOVA, etc, and use that too look at a data set
- this is reverse of what we usually do: in this case we will let the statistics drive the analysis
Dr O has lots of data sets that need to be analyzed
In Agriculture, you don't get setup for your dissertation defense until you have submitted the same thing for publication
in a lot of other programs at other universities, students finish with 4 or 5 publications by graduation
can work in pairs on this project
- key is learning how to put an article together
- learning how to write a technical report in your own field
25 page project will be about 5000 words
Guitar was Dr O's therapist in graduate school
- once he cracked the first song, that opened up everything
Dept of Education released a great CD with lots of data, was longitudinal, etc.
Kirk does mostly between-subjects designs
- also repeated measures / within-subjects design is covered well in another book, Dr O will provide us with a copy of that section
- in education it is a problem that we don't do many repeated measures studies (test students several times, track them over an academic year)
-- just need 40 subjects instead of 300, if you do 4 measures you can
no due dates on assignments because he realized Tuesday about the book problem
next meeting: he will give us deadlines for each project
We have 20 students, before this the largest class he ever had was 12
Our book talks about SASS and SYSDAT, we will skip those parts and just stick with SPSS
it is a good idea to spend a day just screening the data
- run a lot of checks to see if there are outliers, mins and maxes, check to see if data was entered correctly
- there is always going to be errors, often those can be eliminated by running mins and maxes
- descriptive statistics can be helpful here
in SPSS
- always good to know what is happening behind the scenes by looking at the syntax view
SPSS data files
- save .SAV file (data file)
- .SPS file is program syntax
- .SPO file is the print out / result (don't really have to save this)
in projects, Dr O will ask for the syntax for our SPSS calculations
when you are checking to see if assumptions are correct (like homogeneity of variance), we always want to fail to reject (or retain) the null hypothesis
- this is the opposite of hypothesis testing
remember to trim down your SPSS results when you copy and paste into Word
- don't put more info in there than you need!
first 4 projects we will do hand calculations
- last projects we won't because they would take like 2 days to calculate each by hand
the "ocular test" - eyeball it and see how it looks!
Start reading chapter 1 on Tabachnick
2 September 2004
Get the new Green SPSS book and return the yellow one
We don't do experimental design research in education, but many other fields do
about half our course will be on multivariate analysis
- univariate and multivariate analysis is our focus all term
will use some of the exercises in the Kirk book, because our author (Tabachnick) does not have exercises
- Kirk book is available in the LRC to use but not check out
some of these problems here can take 20 pages to hand write out
There are no tests here, just projects that you work on at home (see last page of syllabus)
- you will also do the same on the computer, and explain how the computer results match the hand calculations
we will NOT do written report (like about 12 pages) instead we will have a 1 page impact / implications page
recommendation is to use computer programs and then work it by hand, so you always have the answer
Arturo likes SASS because he likes to write syntax
Graduate pack for SPSS will help us get through the first 4 projects, but probably won't do multivariate
Typical flaws in a study
1- sample size
2- selection of subjects (simple random sampling, etc --> something that gives you a probability sample) - convenience sampling is always a flaw
- power analysis: is it representative / unbiased group? (must have for good statistical estimates, predications, interpretations, and conclusions)
3- instrument: how do I know the instrument is valid, produces valid scores?
-- is it public domain or commercial? (most commercial instruments are good)
--- many public domain instruments are bad
--- generally have a lot of psychometric studies done on commercial instruments
reliability and validity are the psyometric properties that are key
- validity evidence is often more rare
-- reliable is a close group of shots on a target
-- valid result is the group is dead center on the bullseye
when you read the article, you will go back and really focus on the methodology section
- sins of omission and sins of comission
- if you don't find many problems, then you can put a lot of trust in the results
RESEARCH DESIGN is all about how I control for extraneous variation
in quantitative research, we are trying to find/make causal comparisons
- looking for the causes and consequences (that is our goal)
when you find a problem: explain the consequences
ABESCO tends to do real well with fulltext article versions
look for data driven articles
- not just a review
- need to have methodology, results, etc.
We will do our article reviews from journals in our area of interest / study
FOCUS ON THE CONSEQUENCE: WHY YOU THINK IT IS GOOD OR BAD, IN TERMS OF THE METHOD'S EFFECT ON RESULTS
When you do a search on your topic of your dissertation, if you get 3000 results then you may need something else (that subject may have been thoroughly researched)
Don't fall into the trap in your chapter 2 for moving old bones from a cemetary into your dissertation
- it is not about bulk
- tough to be a 70 page wonder
this course is also about writing research pieces
- we are focused on results and interpretation
For every project we do, we have access to the book, the computer, and Dr O (he has the answer for every project we are working on)
- you can always touch base with him and see if you are doing the calculations correctly so far
Get very familiar with one statistical tool: ANCOVA, MANCOVA, etc, and use that too look at a data set
- this is reverse of what we usually do: in this case we will let the statistics drive the analysis
Dr O has lots of data sets that need to be analyzed
In Agriculture, you don't get setup for your dissertation defense until you have submitted the same thing for publication
in a lot of other programs at other universities, students finish with 4 or 5 publications by graduation
can work in pairs on this project
- key is learning how to put an article together
- learning how to write a technical report in your own field
25 page project will be about 5000 words
Guitar was Dr O's therapist in graduate school
- once he cracked the first song, that opened up everything
Dept of Education released a great CD with lots of data, was longitudinal, etc.
Kirk does mostly between-subjects designs
- also repeated measures / within-subjects design is covered well in another book, Dr O will provide us with a copy of that section
- in education it is a problem that we don't do many repeated measures studies (test students several times, track them over an academic year)
-- just need 40 subjects instead of 300, if you do 4 measures you can
no due dates on assignments because he realized Tuesday about the book problem
next meeting: he will give us deadlines for each project
We have 20 students, before this the largest class he ever had was 12
Our book talks about SASS and SYSDAT, we will skip those parts and just stick with SPSS
it is a good idea to spend a day just screening the data
- run a lot of checks to see if there are outliers, mins and maxes, check to see if data was entered correctly
- there is always going to be errors, often those can be eliminated by running mins and maxes
- descriptive statistics can be helpful here
in SPSS
- always good to know what is happening behind the scenes by looking at the syntax view
SPSS data files
- save .SAV file (data file)
- .SPS file is program syntax
- .SPO file is the print out / result (don't really have to save this)
in projects, Dr O will ask for the syntax for our SPSS calculations
when you are checking to see if assumptions are correct (like homogeneity of variance), we always want to fail to reject (or retain) the null hypothesis
- this is the opposite of hypothesis testing
remember to trim down your SPSS results when you copy and paste into Word
- don't put more info in there than you need!
first 4 projects we will do hand calculations
- last projects we won't because they would take like 2 days to calculate each by hand
the "ocular test" - eyeball it and see how it looks!
Start reading chapter 1 on Tabachnick
Posted: Thu - September 2, 2004 at 08:44 PM |
Quick Links
Calendar
Sun | Mon | Tue | Wed | Thu | Fri | Sat |
Categories
Archives
XML/RSS Feed
Comments powered by
Statistics
Total entries in this blog:
Total entries in this category: 12
Published On: Dec 18, 2004 04:57 PM
Total entries in this category: 12
Published On: Dec 18, 2004 04:57 PM
