HomeWork #1 STATISTICAL DESCRIPTION Due, through TurnItIn, by Friday, April 27, 11:59 PM.
Do Not try to submit this a few minutes before the deadline. You may need a few minutes for the submission to go through properly. Make sure you follow and complete each step.
• Look at the following list of variables. All of you will, throughout the quarter, use the following two as your Dependent Variables (DV):
V16 Feeling Thermometer Clinton V17 Feeling Thermometer Trump
Feeling thermometers ask survey respondents to rate candidates on a 0 to 100-degree scale. ‘0’ means, well, they are viewed extremely negatively, ‘100’ extremely positive, with ‘50’ generally indicating neutrality. It was rather amazing to see what percentage of respondents gave Clinton (21.8%) and Trump (31.8%) a rating of ‘0’, and how low the percentage of totality neutral (50°=4.8%/5.1%) and totally positive (100°=5.8°/5.6°) responses were. Comparable ratings for then President Obama (V15) were 15.7% (0), 4.7% (50) and 14.9% (100).
• Now choose one (1) of the following eight Independent Variables (IV) that you will also use throughout the quarter. One of these variables is already binary/dichotomous (2 categories), but others have more than 2 categories, including other feeling thermometers that have a potential 101 (0-100) categories. To place all of you on an even playing field, I have converted (recoded) all these variables into dichotomies. Each is listed at the end of your data file with its original designation (V22A, V81, etc.) supplemented by ‘_30D’ (see the revised CODEBOOK). (1) List the IV that you have chosen.
Demographic IVs V3_30D Hispanic origin or not (already a dichotomy) V4_30D Age (Millennials vs Baby Boomers + (Generation X removed) Issue IVs V22A_30D Favor/oppose ACA (‘neither favor nor opposed’ excluded) V42E_30D Environment increase/decrease funding (‘decrease’% too small—combined
with ‘keep the same’) V47_30D Gun ownership (easier + keep same/ decrease) Feeling thermometers Totally neutral (50°) can be rather substantial here. Only those with a leaning are
included in the first two variables as ‘neutrals’ can be a form of ‘Bradley Effect’ response. The new categories are (1=0 to 49°) (2=51 to 100°). The percentage of negative ratings for SCIENTISTS is extremely low and thus were added to neutrals (0-50°).
V79_30D Illegal Immigrants V81_30D Muslims V81S_30D Scientists (yes, scientists)
• Now create a hypothesis linking your IV with each of your DVs. (2) List that in your submission.
• (3) Offer and list a brief theory sketch as to why you think your two variables should be
related as specified in your hypothesis. What would make you expect the hypothesis to be confirmed more often than not?
You will now begin your foray into data analysis with SPSS (4).
• Using the instructions from the Manual or ‘Trial Run Instructions’
o Open the ANES2016 data set (4a)
• Either by writing out syntax, or using poll-down menus (see Manual)
o WEIGHT the data set using PW2016_FULL (see manual, pp. 14-16) (4b)
—————————————————————————————————————— Weights are often used to adjust certain sampling problems that we find in our data. The ANES researchers use weights to make sure that the sample they draw looks like the data found in the broader CENSUS enumeration on the following: education level, income, gender (again, this is only treated as a binary here). For example, look at the ‘Trial Run’ results. ‘Women’ comprised 52% of the Census data enumeration, but (unweighted valid %) a slightly higher 53.1% in the ANES original sample. To match the ANES sample with the sample, women’s responses have to count a bit less, men’s a bit more. Perhaps a better example would be a series of surveys called the EUROBAROMETER. In order to get a large enough sample from all countries within the EU, the sample size from each country is set at about 1000. Now, that’s fine when we compare countries with each other, but what happens when we want to look at attitudes of everyone within the EU? The 1000 or so from, say, Denmark, winds up counting the same as the 1000 from, say, Germany, even though Germany’s population is about 16 times larger. Before one can look at the attitudes of the entire EU, the responses of Danes have to be weighted down, Germans up until the weighted sample sizes are proportionate to the actual population sizes.
• (4c) Use the MEANS procedure (pp. 59-62), either via syntax or menu to calculate the following
statistics for V16 and, separately, V17 for both of your IV categories (1,2):
COUNT, MEDIAN, MEAN, RANGE, Standard Deviation (STDEV)
You should have four sets of statistics You should have four sets of statistics — 2 categories of
your IV by 2 DVs – to discuss
• (5) Copy all of the relevant tables to a WORD or WORD-translatable file (see pp. 115-117)
• (6) Within that WORD document, describe what each statistic tells you. Be specific. For example, if you come up with a median of ‘57’ for V16, state that “half of all respondents (of
a certain category) rated Secretary Clinton at 57° or higher, half 57° or less. Discuss the
standard deviation, for now, as you would the MAD, and just add ‘with an adjustment for
• (7) Write a thorough discussion of your comparisons – how each of your two IV groups differed on both of your DVs. Again, more is better than less. Make sure to discuss both averages
(median, mean) and variation (standard deviation) in your analysis.
Your grade will be based on the following:
• Correctness of hypothesis (2) • Reasonableness of theory sketch (3) • Correct application of SPSS (4-5) • Correctness and Completeness of Descriptions (6) • Correctness and Completeness of comparison (7) • Professionalism of presentation.
Get started early. The undergraduate assistants can help with SPSS coding. Your section instructors can assist with interpretation. Neither will do your work for you.