Proposal Review

[LaurieLBaker]

Proposal Review

Reviewer: Professor Baker Date: 03/14/2022

Section 1 - Introduction:

The introduction should introduce your general research question and your data (where it came from, how it was collected, what are the cases, what are the variables, etc.).

What is their general research question: We aim to examine what makes a Youtube video popular using a dataset that contains Youtube data. Is the general research question clear? If it is not clear, what questions do you have?

• General question is clear. I think that likes and views are both good metrics for popularity.

Section 2 - Data:

• [X] Is the data in the /data folder?
• [x] Does the README include the dimensions and codebook for the data set?
• [X] Does the proposal include the output of glimpse() or skim() of the data frame.

Data suitability:

• [X] Does the dataset have at least 50 observations and between 10 to 20 variables (exceptions can be made).
• [X] Does the data set include a mix of categorical variables, discrete numerical variables, and continuous numerical variables.
• [X] What variables does the data include (list below): One thing that the team last year struggled with is that many of the explanatory variables are effectively logical (e.g. contains pets or not, etc.). Duration is a continuous variable which could tell you if longer videos are more popular, but you may struggle to get explanatory power in your model (you may have low R-squared values).

  Section 3 - Data analysis plan:

• [X] Does the proposal include outcome (response, Y) and predictor (explanatory, X) variables they will use to answer your question? And/or the comparison groups they will use, if applicable.
• I think there are a couple of interesting things you can do here. You can explore likes or views as a function of the other variables.
• You can look at changes overtime in YouTube videos (do we see steady pet content in terms of when videos were published to YouTube? Or are there peaks and troughs?)
• Older videos might have more likes/views just by nature of being on YouTube for longer. You may want to create a metric of likes/time on air (days). You can calculate this from when the dataset was compiled: date dataset compiled - date video published = number of days on air.

Do the outcome and predictor variables and/or comparison groups make sense in the context of the question? Why or Why not?

• They do, but they could be challenging.
• Some other things you might be interested in looking at is what categories occur together (maybe by doing some cluster analysis). I think also there could be some fun analysis/word clouds you could do from popular videos. There is a great resource of text analysis that you can try out.

• [X] Does the proposal include some very preliminary exploratory data analysis, including some summary statistics and visualizations, along with some explanation on how they help you learn more about your data. (They can add to these later as they work on their project.)

• [X] Does the proposal include the statistical method(s) that they believe will be useful in answering your question(s). (They can update these later as they work on their project.)

• The way your question is framed would lend itself well to a linear model.

• [X] Do they include what results from these specific statistical methods that are needed to support their hypothesized answer?

Reflections

1. What was something you found interesting about the project?

   • It's a really interesting dataset. It will be fun to see what you find.

2. What ideas/feedback do you have for other things they may explore?

   • See above.

3. What kinds of plots should they consider to complete the project goal to create some kind of compelling visualization(s) of this data in R?

   • Time series, heat maps (for categories), word clouds for descriptions, etc.

4. Any additional feedback you'd like to give the other group:

   • I'm interested to know whether the data quality varies. Is it optional to include a description of the YouTube video?
   • I think there will be some fun preprocessing to do with the lubridate package if you want to look at videos of a certain category.
   • You might want to think about different events or things that might explain the rise or popularity of certain categories.

Great work!

[jtam304]

What I found interesting in this project is that their research problem is essentially how someone can get famous with digital media, specifically YouTube. I usually think that the best way to do it is through consistency and quality, but consistency should first be prioritized. I think the other team should look at how consistency can affect a Youtuber's views for a specific video as well as their channel growth. The plots that they should make are definitely scatter plot and maybe pie charts to see in what category specifically to videos get the most views on. Overall, the proposal looks good and they are definitely in the right direction.