IS 372 – Fall 2020 Project 1
This is an open book, open notes, open internet access project. The only rule is that you must do your own work. Make sure you follow this rule to avoid violating academic integrity policy!
For this project, you must follow the scientific method to analyze a dataset of your choice. You can use the ESS data, ICPSR data, or any other dataset from a publically available source. I did a search on Google for “publicly available datasets” and found many. One useful link was Quora: https:
www.quora.com/Where-can-I-find-large-datasets-open-to-the-public, but many others datasets are available.
Once you find an interesting dataset, you should review the metadata, codebook, or any other file that identifies the variables in the dataset. You should then develop an interesting research question. Once you are satisfied with your research question you will need to download the data and analyze it using R, SPSS Statistics, or SAS.
For this project, we are going to keep it simple by following the Scientific Method discussed in class:
1. State a research question
a. A simple example from the ESS data would be; do more trusting people have different TV watching habits than those who are less trusting?
2. Generate theory
a. Here you would do some basic research on the web or by some other means and use it to support your hypotheses. Make sure you cite your references! Do not make unsubstantiated claims!
3. State your hypotheses
a. A simple example based on the research question in 1a would be; Trust and time spent watching TV are positively related.
4. Measure variables
a. Here you need to explain how the variables were measured
5. Graph the data
a. Here you will need to generate a graph that best represents the data and your analysis
i. Descriptive statistics (histogram, etc.)
ii. A graph that supports the results in number 6 (bar graph, line graph, etc)
6. Analyze the data
a. Choose and justify a particular analysis approach (e.g., regression, ANOVA, mean comparison, etc.).
7. Write up the results
a. Concise write-up of your results. Andy Fields provides many examples in the text.
Deliverable
Your deliverable should include your name, the title of your project, an executive summary, the details associated with the steps above, and a
ief conclusion. The file should be a word document with the graphs embedded. Please name your file LastNameFirstNameProject1 Please also submit your R script files and the dataset in Canvas.
The Role of Gender in News Consumption Behavio
(Basic example using R)
Executive Summary
This study seeks to understand how gender influences consumption of news about politics and cu
ent events. The study analyzes ESS data[footnoteRef:1] to answer the Research Question: Do males or females consume more news about politics and cu
ent events? Supporting Hypothesis 1, independent t-tests results show that males were significantly more likely to consume news about politics and cu
ent affairs. However, the effect size associated with this difference was extremely small, and therefore readers should exercise caution when interpreting the significant result as an important difference between the two groups. A conclusion provides a discussion of these results. [1: Add reference to your data here…in this case the link is included in the presentations in Canvas]
Research Question
Do males consume more news about politics and cu
ent events than females?
Theory (this is
ief example only; make sure you provide additional evidence!)
According to the American Press Institute, gender influences news consumption habits[footnoteRef:2]. Specifically, they state that …. And so and so also state that…. Based on this evidence the following Hypothesis is proposed.[footnoteRef:3] [2: https:
www.americanpressinstitute.org/publications
eports/survey-research/news-habits-trust/] [3: You would need to add more justification here. ]
Hypothesis
Hypothesis 1: Males consume more news about politics and cu
ent events than females do.
Variables
Measurement of the variables is as follows:
News Consumption: A continuous scale based upon the following statement: On a typical day, about how much time do you spend watching, reading or listening to news about politics and cu
ent affairs? Please give your answer in hours and minutes.
Gender: Male, Female, or no answe
Graphs
Figure 1 depicts the histogram for the news consumption variable. The histogram shows a positively skewed distribution. The extremely skewed distribution is due to R’s inclusion of the missing values coded as 7777, 8888, and 9999. Absent these missing values, the highest response is 1440, which reflects 14 hours and 40 minutes (please refer to the codebook for information on how this variable was measured). Figure 2 depicts the graph after restricting the news consumption variable to a maximum of 14 hours and 40 minutes.
Figure 1: Histogram for News Consumption Variable
Figure 2: Histogram for News Consumption Variable without Missing Values
Note that the skewness of the distribution remains because of a few outliers roughly above 2 hours and 50 minutes a day. Figure 3 depicts the histogram when restricting the maximum news consumption to 250. As can be seen, setting the maximum to 250 has reduced the skewness of the distribution. However, it is important to note that we have changed our population to those who consume 2 hours and 50 minutes or less of news related content.
Figure 3: Histogram for News Consumption Variable – Maximum 2 hours 50 minutes
Although both graphs reveal a non-normal distribution of the raw data, it is important to remember that the sampling distribution is what matters. For large samples such as the one in the cu
ent study, we expect the sampling distribution to be normal (Field 2018).
A box-plot comparing news consumption across Gender allows us to explore the news consumption data further. As Figure 4 shows, the box-plot identifies a large number of potential outliers for both groups. However, given the nature of the question (news consumption in hours and minutes), it is entirely possible that the values reflect actual news consumption behavior. Further, the measure reflects hours and minutes, yet the data reflect a continuous scale (250 is actually 170 minutes). Thus, the differences among the values are exaggerated. Future studies should convert all news consumption scores to a minutes scale.
Figure 4: Boxplot for news consumption across gende
Analysis
Because Hypothesis 1 compares mean levels of hourly news consumption across gender, the study utilizes an independent t-test.
Results
Hypothesis 1, which states that males consume more news about politics and cu
ent events than females was supported; males (M=64.28) report consuming more news content than females (M=57.64), MD = 6.642, CI 5.564 – 7.719, t(32952) = 12.084, p = .000. However, the effect size for the relationship was small (Cohen’s d = .13). Large sample sizes can cause spurious relationships to be significant, and in this case, the sample size was in excess of 30,000. Based upon Cohen’s d, (.2 = small effect, .5 = medium effect, .8 = large effect) it appears that the difference between the two groups is small. Figure 5 graphically depicts the results:
Figure 5: Graphical results of news consumption across gende
Conclusion
Based on information gathered from a number of sources including the American Press Institute, predictions were that males would report consuming more politics and cu
ent affairs related news than would females. Although significance tests supported Hypothesis 1, the effect size was small. Based upon the combined interpretation, it appears that males and females consume political and cu
ent affairs news related content relatively equally.
References
Field, A XXXXXXXXXXDiscovering Statistics Using IBM SPSS Statistics, 5th Edition Andy Field SAGE, ISBN-13: XXXXXXXXXX
Possible grade improving enhancements
1. Using more sophisticated analysis such as regression, or one way ANOVA
2. Including multiple hypotheses and multiple analyses
3. Combining hypotheses and analyses types
4. Including additional graphs
5. Citing several sources to support your hypotheses
6. Using non-parametric tests to account for outliers (e.g., Spearman r, Kendall’s Tau)
7. Using robust tests such as bootstrapping