I want this recode or computed variable
Need help with a derived variable? Post here.
posted by Paul Gronke at 11/06/2006 10:54:00 PM
A forum for online discussion of class materials, class issues, and other political stuff related to my courses.
posted by Paul Gronke at 11/06/2006 10:54:00 PM
35 Comments:
We need some ideas on how to code religious sentiments, evangelicalism, etc.
Anyone have thoughts?
I need an efficacy recode for the cumulative study. I found the variables that correspond to the 2004 study (v5201 - v609; v5202 - v613; v5203 - v622; v5204 - v624) but I'm not sure how to do the recode. Help?
I'd like a variable for the states that had gay marriage amendments on their ballots - Arkansas, Georgia, Kentucky, Michigan, Mississippi, Montana, North Dakota, Ohio, Oklahoma, Oregon and Utah. Thanks.
I need the strength of preference feeling thermometer variable "folded" to determine how much a person cares.
I believe the code is v3204
Thanks,
mcgarry
My topic is strength of pref vs. care who wins vs. activity. I have one variable for strength, and 2 variables for political activity that you've made in the past, and i have four variables for care. Should I condense my care variables into 1 index in the same way that activity is?
Hi-
Could you create a 2000 version of the p2102-child variable? I would like to be able to compare the two elections.
Thanks-
Lisa
Bailey,
I think we handled you (complex though it was).
Others may want to know what we did: we ended up creating a series of temporary variables: bailey-eff1, bailey-eff2, etc, using the compute facility, then added these up and divided to get an efficacy measure.
Zack
Here is how to do this. Post again if you can't do it.
Go into the "create variables" section and select "compute a new variable"
Paste this into the "Expression" section and see if it works. If you amke a mistake and have to do it over, make sure you click "replace that variable if it already exists":
IF (v1202 eq 5 or
v79 eq 13 or
v79 eq 21 or
v79 eq 26 or
v79 eq 28 or
v79 eq 30 or
v79 eq 38 or
v79 eq 39 or
v79 eq 40 or
v79 eq 41 or
v79 eq 49)
zack_state=1
ELSE
zack_state=0
Natalie,
Go into the 2004 study, select "create new variables," then select "recode variables" and do this:
Name for new var: mcgarr_pref
Variables to use: v3204
Now fill in the three columns this way
(value / label / var1 )
1 Strong 25,26,27,28
0 NotStrong 11,12,13,14
Lisa,
Go into the 2000 study, choose "create a new variable", choose "compute" and paste this into the expression window:
==== cut here ====
if ((v1586 lt 7) and (v1587 lt 7) and (v1588 lt 7) and (v1589 lt 7)) p210-child=((v1586)+(v1588)+((v1587-6)*(-1))+(v1589))/4
=====cut here ==============
When this is done, run a correlation matrix of v1586, v1587, v1588, v1589, and p210-child. There should be positive correlations betweenv1586, v1588, and v1589, and a negative correlation between 1587 and the index.
I tried to run a regression, and it gave me this:
ERROR: Cannot calculate the regression. (The specified variables may not have any valid cases in common.)
What does that mean?
Hope you had a good Thanksgiving. When I copied & pasted the code, I got the following message:
ERROR: Parentheses do not match in compute expression line: IF (v1202 eq 5 or
After playing around with the parentheses, the program told me there isn't a v79 in the 2004 study. I changed all of the v79's to v1202's, but about half of the states weren't there, so I had to delete them. The variable I made looks like this:
IF (v1202 eq 5 or v1202 eq 13 or v1202 eq 26 or v1202 eq 39 or v1202 eq 41 or v1202 eq 49)
zack_state=1
ELSE
zack_state=0
Is this OK?
Also, is it necessary to run a bivariate analysis? Can we just do several multivariate analyses?
I need to recode variable v3294 (R Income Level) for use in my regressional analysis. However, the incremental values of the survey data do not allow for me to recode them as equal intervals (and are not equal themselves). Any suggestions?
sorry, the previous comment is mine.
One more question: for my paper I am looking at activism and perceived political efficacy of 18-24 year olds; Paul, you have recoded variables (p210-activism and p210-efficacy, respectively), but neither of these give a description of their component parts. Can you illuminate the innerworkings of these recoded variables for me?
the description of p210-efficacy doesn't list the variables that it uses. are they 5021-5024 in 2004?
thanks
hmm... same number of valid cases. probably not a coincidence.
Michelle,
This means you put in variables for which there are no responses. Most commonly, it means you included two variables that are mutually exclusive (for example, suppose you included a variable that asked about Hispanic descent that was only asked of Hispanics, but you also included a questtion asked only of Whites).
What was the regression?
Zack,
Make it only LOOOOONG line.
Zack,
That looks OK. Stupid that it doesn't run if the states aren't there--that should just pass through the command.
Michelle,
i would like a bivariate analysis. Part of the assignment is to learn how to compare a bivariate analysis to a multivariate analysis.
xeno,
My mistake. I recoded this variable because I screwed up the initial coding, remember? I divided incorrectly and it was -.5 to .5. So I just did a mathematical transformation.
you are right, those are the input variables.
Matt,
Why would you want to recode them to equal intervals? Income is not equally distributed--it is a "fat tailed" distribution, meaning that is has a fat right tail, and it is non-normal. In a case like this, it's much better to categorize the variable so that you have roughly equal numbers of respondents in each category (hence the reason that income is typically reported in quintiles or quartiles).
matt
I think p210-activism is reported (do a view and see if the component parts are reported). Efficacy is a scale of the four efficacy items as we discussed in class. I recoded it so that it now runs 0-1.
I was wondering about the recoded variable "p210-disaffect."
In the codebook, the description is:
__________________
Created by RECODE version 2.0
on Nov 23, 2005 (Wed 12:07 AM PST)
Input1: v5242 label: Q7. CSES Makes a difference who is in power
Input2: v5243 label: Q8. CSES Who people vote for makes a difference
From study: /D3/NES2004public
Output Input1 Input2
.00 5 1
.25 4 2
.50 3 3
.75 2 4
1.00 1 5
All other combinations of input variables
were converted to system-missing
______________
When I ran it, though, the results range from "not powerful" to "powerful." How do I figure out how this variable is supposed to be read?
Angelique,
This is a "disaffection" variable. Essentially, high values (1) means that someone said they don't think it really matters who is in power, and that who you vote for really does NOT make a difference.
Low scores mean the individual disagrees with both of these statements.
I want to see how many people that were active in their community were also active in the election. The variables are
p210-activism
p210-electactive
I've run a frequency table on the two, but I'm having a hard time figuring out the table. Is the frequency table what I want? If it is, how should I read it?
I also want to see how many people that were electactive are also generally interested in political campaigns (v5001), would i go about doing this in a similar way?
Thanks,
McGarry
v5008 in the post-election survey asks if the respondent was contacted by a political party prior to the election. There's a clear correlation that people who were contacted are much more likely to have voted. Is there any variable to determine whether or not that contact was (self-identified as) influential, or why the parties contacted the voter (for example, if the respondent was contacted due to prior involvement/interest)?
Is there any way to get recoded variables for activism and mobilization in the 2000 NES?
Angelique,
No, there is no way to tell. You have to simply state that they were contacted. And yes, it makes a huge difference. Try running turnout x educ, then turnout x educ then controlling for contact.
interesting, eh?
Matt,
Can you be more specific?
Matt,
p210-electactive
p210-partymobil
p210-activism
all coded for 2000 study.
Hello my friends! Who are you!?
Post a Comment
<< Home