https://dacss.github.io/DACSS_601_Summer2023_Sec1/posts/Homework3_Xinpeng_Liu.html

[giscus[bot]]

https://dacss.github.io/DACSS_601_Summer2023_Sec1/posts/Homework3_Xinpeng_Liu.html

Study for Hotel bookings data set

https://dacss.github.io/DACSS_601_Summer2023_Sec1/posts/Homework3_Xinpeng_Liu.html

[sp-conway]

Hi Xinpeng,

I took notes as I read, which are posted below. Please let me know if you have any questions. In general, I think your project would benefit from more exploratory data analysis, a little bit more explanation of the figures you do include, and cleaning up the graphs to make them easier to read. However, you are making great progress.

• Nice description of the dataset. Please make sure to mention where you found it (i.e., cite the data source).
• Nice job describing the data mutation.
• Nice research questions.
• Booking patterns
  • Number of bookings per month per year
    • Rather than have a facetted plot, I would plot all dates on the x axis (maybe draw a vertical line for the beginning of each year?
      • Also fix the x axis labels (probably need to rotate them using theme(element_text(hjust=0.5))
      • If you do this, you will no longer need the legend. Though if you keep the legend (and I recommend against it) change the numbers to month names
  • Average lead time per month per year
    • Same comments as before.
• Guest preferences
  • Double check the source of the data on gaggle to see if there is a description of the room type. I can’t remember if there is.
  • Room type vs. number of adults
    • I don’t think a box plot makes sense here. I think a simple summary table showing the majority of rooms are booked for just a few people, regardless of room type.
  • Room type vs. number of children
    • Same as above comment
  • Room type vs. number of babies
• Cancellations
  • Cancellation rates by lead time
    • Just plot the proportion of cancellations. Not the proportion of cancellations and the proportion of non-cancellations.
  • Cancellation rates by room type
    • Same as above
  • Cancellation rates by repeated guest status
    • Same as above
    • Relabel variable on x axis. Use recode() or mutate() and case_when() to change 0 and 1 to more informative values.
      • Also relabel variable in the legend
  • I would strongly recommend against using a logistic regression in this way for this project. Unless you have specific hypotheses to test, I would focus on exploratory data analysis.
    • See the rubric for the final project
• Special requests
  • Histogram of special requests frequency
    • Nice
  • Cancellation rates by special requests
    • Only plot one proportion
      • See above comments
    • Is this plot conditioned on having made a special request? That is, given that you made a special request, did you later cancel? I would make sure this is clear to the reader.
• Pricing strategies
  • ADR by month
    • Relabel month
    • Use alpha to change transparency. See my post in random in slack
  • Adr by room type
    • Maybe remove outliers
      • It’s hard to see a trend with that one outlier compressing the box plots