A great row-wise mapping function family called slider which could be applied in high-performance window functions.
Plans, Goals, Interests
Some key parts that I want to achieve in the next project:
A shiny app that follows package development practice
More than lines and dots:
The dashboard should not be stacked with plots
For each type of plot presenting, ask ourselves to categorize the information that this plot is about to deliver. And generate narratives that correspond to each of the segments and categories.
Create more intermediate-level, interactive, data-driven, customized narratives
Less proactive tweaking
Create something that people won't need to proactively open, but it's always there
mailing with directions and summarize
Keep our users inside the application: Communication
Allow comment under visualizations to bring in more discussions and keep track of the iterative insights to make the comments an integral part of the plots.
The insights from our users can also be seen as important textual data. It might be reproducible, but the cost may be expensive, and the quality may vary. We should treat plots like objects with attributes that keep track of the comments.
library(tibble)
library(ggplot2)
plot <- tibble(
title = character(),
plt_obj = list(ggplot()),
date = date(),
comment = tibble(
date = date(),
content = character(),
publisher = character()
)
)
str(plot)
tibble [0 x 4] (S3: tbl_df/tbl/data.frame)
$ title : chr(0)
$ plt_obj: list()
$ date : chr(0)
$ comment: tibble [0 x 3] (S3: tbl_df/tbl/data.frame)
..$ date : chr(0)
..$ content : chr(0)
..$ publisher: chr(0)
What to do:
* Find a problem, a dataset, or some datasets
* It will be helpful for Wil to assign me more tasks to begin with. It's not that I am not confident. I don't think I am there yet to start development on my own. So task assignments would be a great place to start.
# Meeting Follow-ups
## Goal of analytics development
Make our development as simple as possible. Be sure of the core functionality of our work and try not to reproduce work that has already been done nicely elsewhere.
Feedback on the comment section idea: Wil and I both agree this would be a good idea. But the team has been trying many ways to help more people engage in conversation regarding analytics topics, but the progress is not significant so far. Why is this happening?
* Many people in the organization have more than ten years of experience in this business, so they would rather start a phone call to people with interests than reply to the conversation threads.
* Our tools might not be closer enough to where the decision is made. Wil mentioned that the buyers or sales would usually make their final decisions with Oracle opened, which explained why we want to make our tools light-weighted and provide more APIs to be applied elsewhere.
To conclude this part of the discussion, maybe our tools could be more "competitive" and more integral to our user's work routine. Not a NONO to the comment section and bring in more conversations to the visualizations, but the feedback and documentation culture still need to be carefully cultivated. It is still valuable for us to collect feedback and insights from our users. They do not have to be plenty. Most of the time, one or two sentences about the visualization from our expertise would be crucial to making decisions.
## A forecasting tools
The team has been getting many forecasting requests, according to Wil. We want to build a simple tool for users to upload their data and perform time-series analysis.
Potential functionalities include:
* Seasonality detection
* Trend detection
* Data missing information
* Predictions
Something to keep in mind:
* Do not expect your users to acquire data literacy or capability to extract what they need from stats plots.
* Provide them with next-step guidelines based on their data and what we generated.
* Plots might not be more useful or persuasive than convincing narratives.
* Automation, don't repeat yourself. Use **Shiny Modules** to pile up applications.
* Go further, don't just provide them with what they asked. For example, a user asked for a prediction. We provide them with predictions, expect, and prepare to cover most of the scenarios based on prediction intervals.
## Research on the Bid-response Model for customized pricing
* The [Bid-response model](https://github.com/WorthingtonIndustriesAnalytics/wipeo/blob/master/vignettes/research/Bid-Response_Models_for_Customized_Pricing.pdf) was introduced in 2007 by Agrawal and Ferguson in Georgia Insitute of Technology.
* A [Bayesian model ](https://github.com/WorthingtonIndustriesAnalytics/wipeo/blob/master/vignettes/research/Optimal_Customized_Pricing_in_Competitive_Settings.pdf) was carried by them in 2006.
2021-05-26
Thoughts & Plans Before Meeting
Work Summarize
wicast,wipeotsibbleobject data structurePlans, Goals, Interests
Some key parts that I want to achieve in the next project:
plot <- tibble( title = character(), plt_obj = list(ggplot()), date = date(), comment = tibble( date = date(), content = character(), publisher = character() ) ) str(plot) tibble [0 x 4] (S3: tbl_df/tbl/data.frame) $ title : chr(0) $ plt_obj: list() $ date : chr(0) $ comment: tibble [0 x 3] (S3: tbl_df/tbl/data.frame) ..$ date : chr(0) ..$ content : chr(0) ..$ publisher: chr(0)