Final Project Questions

Someone asked:

1. Should the length of our write-up be comparable to the paragraphs we wrote in the homework assignments?

2. In reporting statistics, do you use the 0 in front of the decimal rule for that particular case or for any case in which the value could exceed 1? For example, with means when obviously most of the time the means do exceed 1, but in certain circumstances the mean may be less than 1? Would you put the zero in front of the decimal in the latter example? Does that make sense?

Homogeneity of Variance

So, I ran a Factorial RG, Fixed Effects and found a significant interaction and no significant main effects but I have a substantial violation of homogeneity of variance. When I computed the Fmax, it came out to 15.54. First of all, is this considered moderate or severe (which would impact my significance at p = .016)? Secondly, in looking at the issue of nonorthogonality, since my highest order effect, my 2-way interaction, is significant but none of the others are, can I ignore the issue of heterogeneity, or is the most substantial issue going to be in my comparison of marginal means?

Missing definition

I don't know if I saw this correctly, however there seems to be a missing variable in our paper description that is in the data set, entitled momid. The variable menu describes this as identification mom, but I don't quite know what this means. Would you be willing to describe this briefly? Thanks
Kris

Reverse Scoring

Someone wrote:

I am a little confused on the direction I should go involving transformations of the DV. I have evaluated my DV (self-esteem) and the histogram reveals a moderate negative skew. In the book, as well as in your handouts, the approach instructs me to reverse score (reflect) the raw scores and then to conduct a square root transformation. I am not clear on how to reverse score my DV, nor am I certain on how to execute the transformation in SPSS. From the "Transform" menu where do I go next? The book gives instructions using syntax. I do not know how to get to the syntax command center.

The handouts state that with relatively equal cell sizes, no outliers, tow tailed tests, and df (error) > 20, ANOVA is robust. My cell sizes are not equal and there are a few outliers. Are we to do Cook's D or leverage tests for these outliers?

insignificance!!!!

so...for my final project, i decided to go with a randomized group ancova. after running the analyses, i found that my CV significantly effected my DV. however, the main effects of my IVs and interaction effect i was hoping for turned out to be insignificant!! should i report this for my final project, or should i just pick different variables and maybe a different anova that will render significant effects?

Exam 3 Review Sheets

Your individual review sheet (with items missed on previous exams) is available for pickup at the front desk. The generic version is below.

The final exam is cumulative, but weighted toward the final third of the course.
Be prepared to interpret printouts from one-way, factorial, repeated measures, and ANCOVA designs. Be prepared to interpret printouts from both fixed effects and random effects models.
Be very clear about the descriptive statistics that should be reported for each analysis of variance design.
Be very clear about the inferential statistics and effect size indicators that should be reported for each analysis of variance design.
Know the assumptions of ANOVA, ANCOVA, and Repeated Measures designs. Know how they are tested, and what strategies should be taken if these assumptions are violated.
Know how to compute the degrees of freedom for the effect portion of the IVs (both main and interaction effects). Understand the differences in how degrees of freedom for error terms are computed across randomized group and repeated measure designs and across random effects and fixed effects designs (you will not be required to compute these).
Know the primary advantages and disadvantages of one-way randomized group models, factorial randomized group models, repeated measures models, mixed designs (repeated measures and randomized groups), ANCOVA, Random Effects designs, and Latin Squares designs.
Know how models that mix random effects and fixed effects behave. Understand the similarities and differences between such a mixed factorial model and a randomized effects factorial model.
Know the applications and limitations of screening designs. Recognize common types of screening designs.
Know when it is necessary to use a post hoc comparison (a/k/a multiple comparison procedure or MCP). Understand issues of power and Type I errors associated with MCPs.
Understand sums of squares, mean squares, and their relation to the F ratio.
Be able to explain the relation between error df and power to detect a significant difference. Understand the relation between error df and F, p2, and p.
Know when to use Type I, Type II, and Type III sums of squares.

SAVE GAS & MONEY WITH FREE TRANSPORTATION!

All members of the Fuller community (students, faculty, and staff) can ride any Pasadena ARTS bus FREE for the entire month of April. Just board the bus, show your Fuller ID card, and enjoy the FREE ride. Staff can pick-up ID cards at the Auxiliary Services office.

All ARTS buses now have bike racks. Please contact Auxiliary Services for more information regarding bus schedules, routes, stops, and maps.

You can also access ARTS information at:

www.cityofpasadena.net/artsbus

ARTS bus stops directly serving Fuller are at Green/Los Robles and Colorado/Oakland.

Route 10 is an easy connection to Gold Line Stations:

-Allen Station

-Memorial Park Station

-Del Mar Station

Route 10 is also an easy connection to:

-Old Pasadena

-Playhouse District

-South Lake

Route 40 travels along Walnut, Los Robles, and Villa

-East Pasadena (Rosemead & Foothill)

-Sierra Madre Villa Gold Line Station (20 minute ride from Fuller)

Bus Routes on Lake Avenue:

-ARTS Route 20

-Metro Route 180

-Metro Route 485

Missing values

So, here's my problem...well, my ANOVA problem anyway. I am looking at trying to exclude cases from my data set and I am able to successfully exclude cases under one criteria (those who's parents are widowed, all of which apparently are dads which helps my analysis) but I would also like to exclude anyone who did not put a score for dadinv as my hypotheses depend upon some level of dadinv. Since the cases that I am looking at which do not have a value for dadinv also don't have values for many of the other dad categories, I thought this would be a reasonable solution because my hypothesis is the relationship between the ratings on the other dad categories and self-esteem. 
SO...can you help me fix this issue in SPSS?
Here is what I have done:  in "Select Cases" I chose the option to select cases "If..." and did pmstatus less than 3.1. What is next if I want to also exclude those cases who don't have a value for dadinv?
Thanks, Kris

POST HOC for ONE-WAY ANCOVA

I'm doing a one-way ANCOVA, with my IV having 3 levels. I'm already done with the analysis but got stuck with Post-Hoc since one cannot use the Post-Hoc function when a covariate is introduced to the analysis. I've played around with how to go about it but have consistently hit a blank wall, please help me....Hehe. God bless.

- Neil C.

TRANSFORMED SCORES, MISSING SCORES and Writing Up the Results

Hi,

I'm doing an ANCOVA and found out that I need to transform my CV due to the presence of several outliers, I was just wondering how the write-up would look live. Would it still be effect of IV on DV, covarying out CV? or IV on DV, covarying out the square root of the CV. Also, at which part of the write-up should I mention the transformation? At the start? Or writing down the results and inferential stats?

Also, due to missing scores, I will need to eliminate some cases, I'm guessing I need to mention this again in the Results section but I am not sure where? Should I mention at the start that X number of cases were not analyzed due to missing data?

Thanks. God bless.

- Neil C.

Where did the term Latin square come from?

If that question piqued your interest, you might want to follow up at
http://en.wikipedia.org/wiki/Latin_squares

But the short answer is that the mathematician who is credited with the first academic work with Latin squares used Greek and Latin characters in the squares, and that he called them Graeco-Latin Squares (or Latin squares when only Latin characters were used in the squares).


(Apparently, Latin squares draw from Magic Squares, a very old Islamic puzzle game in which the numbers 1-9 are arranged in rows/columns of three in which each number is used once and only once and the diagonals must equal 15. And you thought sudoku was a new-ish game!)

Number of Analyses Needed for Final Project

Is there a minimum number of analyses techniques you want us to run besides one Analysis of Variance Technique such as regression or t-tests? For example, if i decided that a factorial anova is the best suitable analysis technique, would i need to run another to verify or confirm my results from the factorial anova? 

Final Project

You should have received an email with the data set and instructions for your final project. (The instructions will also be provided in hardcopy in class.)

If you did not receive the data set, please email me directly so that I can resend it to you.

As a reminder, you may (and in fact are encouraged to) ask questions on the blog about this assignment. You may not work together on this assignment.

Exam 2

All of your exams are at the front desk. The average grade across both sections was a B+.

If you have questions about anything on the test or if your score is lower than you are hoping for on Exam 3, please sign up for a time to come see me.