Transformations HW

Hi - have had a couple of questions about the hw for Tuesday, so I thought I'd clarify:

-Run 3-4 models, using a non-normal outcome that has been transformed 3 different ways (one model though should be with the untransformed version)
-Include covariates in these 3-4 models
-Examine residuals for each of the models.
-Table of the 3-4 models
-Write up - what you're presenting; describe the results including r2 and coefs; describe residuals- mean, min,max, distributions.

missing data for this week's ANCOVA hw

A couple of folks asked about how to treat missing data in this week's hw. Here is Doug's response:
"For the assignments it's up to you. If you want to just keep using pairwise deletion in the regressions, that's fine. We'll talk later about what to do for the final paper."

Centering syntax

Hi - a couple of people have asked about syntax to center a variable. Here it is - you're just creating a new centered variable by subtracting the mean - so that 0 is the mean for the centered variable. This makes interpreting the interaction effects easier (see posts on interactions below...).
-------------------------------------------------------------------------------
desc var=p1ageent .
EXECUTE .

***from the output from above, find the mean and plug it in here, in place of 65.93.

COMPUTE age_cent = p1ageent - 65.93 .
VARIABLE LABELS age_cent 'P1AGEENT centered around its mean' .
EXECUTE .

-------------------------------------------------------------------------------

Missing Data HW

For this assignment use the missing dataset that's posted on classweb instead of the usual ECLS-K dataset.

Explore the randomness of missing data on family income (WKINCOME). Employ both t-test and chi-square analyses to determine which child characteristics are associated with missing data on WKINCOME. For this assignment, focus on race (RACE5), gender (FEMALE) and mom’s age (P1HMAGE). What are the implications of these missing data for analyses involving children’s economic backgrounds?

But also, include 1 more continuous variable and 1 more dummy/categorical variable of your choosing.

Don't write a paper this week - just write two paragraphs, one on the chi-squares that you run and one on the t-tests.

Also hand in your output and syntax as usual.

Welcome Back

Hi guys -

I'm not sure how useful this blog is, but I thought I'd start it up again and see if folks access it, post comments, etc.

Also, I'll be on the 2nd floor of the library on Monday, 2-4pm if you want to go over questions in person...

Interactions, what a great way to start the new semester!! :)

Megan

Interaction Assignment

Using the dummy variable indicating that the child lives in a non-two-parent family (SINGPAR, 1=single-parent or other adult guardian family; 0=two-parent family) and the continuous SES measure (WKSESL), create a SINGPAR*SES interaction term. Employ multiple regression to explore whether the relationship between SES and entering literacy ability (C1RSCALE) depends on family structure (and whether the relationship between family structure and entering literacy ability depends on SES). Create an appropriate regression table.

HW Due November 18th

Conduct three bivariate regressions using the same continuous outcome, but three different continuous or dichotomous predictors. To what extent does the explanatory power of your predictors differ? Be sure to discuss your findings in terms of how a one-unit change in X influences Y. Importantly, in terms of each of your predictors, what is a “unit”? Select one predictor and create a standard regression table.