week 6 Discussion
Topic: Discuss a point you found interesting from this week’s reading(s)and/or video(s).
Requirements
Make at least one explicit reference to something you learned in the reading or the video, with an inlinecitation of that reading/video
If you’d like to relate this week’s materials to the global pandemic, or discuss course relevant topics related to data and/or decision making in the global pandemic, this is a space you can do so.
Minimum250 words. Word count does not include references.
At least one in line citation and reference list in APA format
Due Saturday 11:59p
Secondary Posts (2 posts,25% of grade)
Topic: Responses to other posts, or responses to responses to your post
Requirements
Min. 50 words.
Word count does not include references.
Data Driven Decision Making
Week 6:
Follow up decisions
Statistical Review
1
Statistical (& Data) Review
Validating the approach taken for making a smart decision
The following is based on Randy Bartletts book: A Practitioners Guide to Business Analytics, Section 9 Statistical ReviewAct V
Code Review
One or more engineers (reviewers) reviews a specified portion of the another engineers code
Goals (from Wikipedia):
Better Quality Code
Find defects
Learning/knowledge transfer
Increase sense of mutual responsibility
Finding better solutions
Complying to QA guidelines
Wikipedia contributors. (2020, March 28). Code review. InWikipedia, The Free Encyclopedia. Retrieved 21:22, June 24, 2020, fromhttps://en.wikipedia.org/w/index.php?title=Code_review&oldid=947848980
Statistical Review Purpose and Scope
Purpose
Verify the integrity of the decision or analysis and, potentially, to find alternatives
Scope
Shaped by what makes sense within the confines of our resources (including time), the complexity of the work, and our projected benefits.
Modified from Uwe Hohgrawe / Bartlett textbook
Statistical Review – Benefits
All important analytics-based decisions merit review. Statistical Review is a major force in continual improvement. The benefits of Statistical Review are both immediate and long-term. In addition to continually raising our analytics aptitude, Statistical Review provides numerous downstream benefits such as:
Improves decision making.
Ensures the rigor of the results, thereby enhancing reliability.
Reveals insight into the problemeven a partial review can reveal insight, which can be leveraged to find better solutions.
Provides on-the-job training, hones analytics professionals.
Fosters collaboration between reviewers and those reviewed.
Protects the findings from political aspirations.
Sterilizes one source of political growth.
Eradicates expensive and dangerous fairy tales about how to run the company.
Discourages negligence, charlatans, and counterfeit analysis.
Encourages speedier execution.
Uwe Hohgrawe / Bartlett textbook
Statistical Review Scope checklist
Here is a typical checklist tailored for conducting a particular review:
Purpose of Decision/Analysis
Thoroughness of Review
Statistical Qualifications
Underlying Assumptions
Analysis Structure
Statistical Diagnostics
Alternative Solutions
Timeliness, Client Expectation, Accuracy, Reliability, and CostBSP (Best Statistical Practice) List
Decision Results
Recommendations for Future Enhancements
Response
Uwe Hohgrawe / Bartlett textbook
Review should be an upbeat, collegial opportunity to encourage professional norms, an opportunity for nurturing technical competence in other. We should find out what went right and suggest what we can do better. Review is a development opportunity
– Randy Bartlett, A Practioners Guide to Business Analytics
Reviews often happen as postmortems when something goes wrong
Better: do them on a regular basis and in a positive supportive culture
Uwe Hohgrawe / Bartlett textbook
Follow up decisions
The decisions never stop
The decisions never stop
We are constantly making decisions
35,000/day is frequently cited but most likely proverbial*
All parts of our lives
Suck time and energy out of us
Hulu Live TV cost went up $10/mo! Still worth it?
* Stackexchange. (2018).Basis for “we make 35,000 decisions a day” statistic.Psychology & Neuroscience Stack Exchange. Retrieved 15 August 2018, from https://psychology.stackexchange.com/questions/17182/basis-for-we-make-35-000-decisions-a-day-statistic
My Cable/Internet Decision Making Spreadsheet
Analysis Paralysis / Decision Paralysis
Youll see I wear only gray or blue suits. Im trying to pare down decisions. I dont want to make decisions about what Im eating or wearing. Because I have too many other decisions to make.
– President Barack Obama, when in office
xkcd: Decision Paralysis. (2018).Xkcd.com. Retrieved 15 August 2018, from https://xkcd.com/1801/
Always Wear The Same Suit: Obamas Presidential Productivity Secrets. (2014).Fast Company. Retrieved 15 August 2018, from https://www.fastcompany.com/3026265/always-wear-the-same-suit-obamas-presidential-productivity-secrets
Process of Decision Making
1. Define Objective
2. Reduce ambiguity and risk
Collect data
Analyze data
3. Make a choice
4. Execute
5. Measure and adjust according
Modified from: Figliuolo, Mike. (2018).Decision Making Strategies. Retrieved 8 October 2018,
from https://www.lynda.com/Business-Skills-tutorials/Defining-decision-making/186697/373496-4.html?org=neu.edu
Path is frequently non-linear
May skip steps (e.g. data collection)
Many subdecisions in the path
Deciding on a goal
Deciding on what data to collect/use
Etc Yufei Sun
Week 6
COLLAPSE
Top of Form
Regarding to week 6 material, I am very interested in the topic of follow up decisions. As the professor said, we are constantly making decisions. we make 35,000 decisions a day (Stackexchange, 2018). We might make thousands ofdecisions to reach the final decision. In the process, there will be lots of problems come out, we have to make decisions to deal with them. For example, when we prepared to buy stuff, we have to consider if the price is suitable. And we have to decide on the brand that each brand has their own advantages and disadvantages. In addition, we might have to decide the size and where to place it. moreover, we might consider if we could save money to buy other stuff. To get the stuff we will face lots of decisions because there are lots of potential problems we have to consider.
There are lots of benefits that follow-up decisions can bring to us. On of the benefit is the follow-up decisions will reduce the risk of the final decision. The more problems you come up with, the more decisions you will make, and then the less risk you will have. I believe each company does the same thing just like our project. When we get the final decision, there will be some follow-up decisions. Thus, it is necessary to resolve those problems that will make sure our final result nothing falls.
This is our final week,I am very glad to have the same class with all you guys.Wish all you best.
Thanks
Reference
Stackexchange. (2018). Basis for “we make 35,000 decisions a day” statistic. Psychology & Neuroscience Stack Exchange. Retrieved 15 August 2018, from https://psychology.stackexchange.com/questions/17182/basis-for-we-make-35-000-decisions-a-day-statisti
Bottom of Form
Mengqiu Yao
week 6 discussion
COLLAPSE
Top of Form
Though Google Flu Trends project was a failure, we shouldnt deny the potential contribution of big data to public health. The value of the data held by entities like Google is almost limitless, if used correctly (Lazer. D & Kennedy. R, 2018). The theory behind GFT is if there are large number of searches related to influenza in certain area in a period, then theres a high possibility that theres a corresponding influence population in the area. System will trigger early public health warning to relevant departments. The GFT failed because it overestimated the incidence of illness and sometimes doubled the data from CDC. The data is powered by Googles rich search data in the flu season. The search results are highly related to the self-estimated flu symptom from the user population. Some people searched for related flu information because they were worried, and GFT may potentially marked these people positive with flu. However, Google may have potentially overestimated flu rate by doing so because they dont have direct evidence to prove the illness. For example, searches for the COVID-19 may surged from February to June only because people want to get relevant information, but it doesnt mean the COVID-19 cases increase sharply in an area at then. Google may have ignored the noise caused by these search results in the prediction model, thus affecting the accuracy of the model. Also, GFT over relied on statistical correlation, and directly replace it with cause and effect between things. I think the reason of the failure is that the analysts didnt figure out the connection between keyboards searching and the spread of influenza, and they failed to find the reasons behind the connection. More and more cases have proven that data is not the more the better. No matter what, I think GFT is a very good attempt if data analysts can learn from it.
Reference:
Lazer, D., & Kennedy, R. (2018, June 06). What We Can Learn From the Epic Failure of Google Flu Trends. Retrieved August 11, 2020, from https://www.wired.com/2015/10/can-learn-epic-failure-google-flu-trends/
Bottom of Form
