ASA Connect

Many important insights here! While many of us have long been cautious of buzzwords like "Big Data", this article details when more data is needed, when it isn't<g class="gr_ gr_45 gr-alert gr_tiny gr_gramm gr_inline_cards gr_run_anim Punctuation replaceWithoutSep" id="45" data-gr-id="45">,</g> and underscores the practices we need to be pursuing. It's good to see people realizing that putting more records in the model just for the sake of more records doesn't automatically make the model better. Maybe it helps, maybe not, maybe it doesn't help enough to be worth the additional effort and resulting delay in taking action on the results. Careful design of experiments, selection of the right modeling methods, and scientific sampling will continue to have a significant impact. Relying on brute force alone, unaided by careful science, will give us brutish models.

------------------------------
David J Corliss, PhD
Analytics Architect / Predictive Analytics
Ford Motor Company
------------------------------

Original Message:
Sent: 12-01-2017 10:00
From: John Major
Subject: "Big data is dead. Data is just data."

This essay by Alexander Thamm provides a perspective I think most statisticians can align with:

https://www.linkedin.com/pulse/big-data-dead-just-regardless-quantity-structure-speed-thamm

 

 

John A. Major, ASA, MAAA

Director of Actuarial Research, Guy Carpenter Analytics

1166 Avenue of the Americas

New York, NY 10036

860-839-9148 (cell)

https://scholar.google.com/citations?user=9b7B9C4AAAAJ&hl=en

 

---

This message is intended only for the use of the addressee and may contain
information that is PRIVILEGED AND CONFIDENTIAL.

If you are not the intended recipient, you are hereby notified that any
dissemination of this communication is strictly prohibited. If you have
received this communication in error, please erase all copies of the message
and its attachments and notify the sender immediately. Thank you.

Original Message------

Many important insights here! While many of us have long been cautious of buzzwords like "Big Data", this article details when more data is needed, when it isn't<g class="gr_ gr_45 gr-alert gr_tiny gr_gramm gr_inline_cards gr_run_anim Punctuation replaceWithoutSep" id="45" data-gr-id="45">,</g> and underscores the practices we need to be pursuing. It's good to see people realizing that putting more records in the model just for the sake of more records doesn't automatically make the model better. Maybe it helps, maybe not, maybe it doesn't help enough to be worth the additional effort and resulting delay in taking action on the results. Careful design of experiments, selection of the right modeling methods, and scientific sampling will continue to have a significant impact. Relying on brute force alone, unaided by careful science, will give us brutish models.

------------------------------
David J Corliss, PhD
Analytics Architect / Predictive Analytics
Ford Motor Company
------------------------------

Original Message:
Sent: 12-01-2017 10:00
From: John Major
Subject: "Big data is dead. Data is just data."

This essay by Alexander Thamm provides a perspective I think most statisticians can align with:

https://www.linkedin.com/pulse/big-data-dead-just-regardless-quantity-structure-speed-thamm

 

 

John A. Major, ASA, MAAA

Director of Actuarial Research, Guy Carpenter Analytics

1166 Avenue of the Americas

New York, NY 10036

860-839-9148 (cell)

https://scholar.google.com/citations?user=9b7B9C4AAAAJ&hl=en

 

---

This message is intended only for the use of the addressee and may contain
information that is PRIVILEGED AND CONFIDENTIAL.

If you are not the intended recipient, you are hereby notified that any
dissemination of this communication is strictly prohibited. If you have
received this communication in error, please erase all copies of the message
and its attachments and notify the sender immediately. Thank you.

Is it possible that "Big Data" was replaced with "Machine Learning" and "Data Science" because those are the types of skills needed to analyze "Big Data"? 

For anyone that thinks data is just data, try working at Ford on autonomous cars or at U of Michigan health system taking signals from medical equipment. Both require taking live data from multiple streams of sensor arrays or arrays of sensors, and determining, "Is this data good?, If so, then what?" Get it right, no one knows or cares and things get better. Get it wrong, people die. 

Having worked with this type of data, it's not uncommon for each sensor to send 100Hz to 1,000Hz of signal data times X number of sensors times Y hours.... I think that gets pretty big. Thankfully, there is Machine Learning.

------------------------------
Andrew Ekstrom

Statistician, Chemist, HPC Abuser;-)
------------------------------

Original Message:
Sent: 12-04-2017 08:54
From: David Corliss
Subject: "Big data is dead. Data is just data."

Many important insights here! While many of us have long been cautious of buzzwords like "Big Data", this article details when more data is needed, when it isn't, and underscores the practices we need to be pursuing. It's good to see people realizing that putting more records in the model just for the sake of more records doesn't automatically make the model better. Maybe it helps, but maybe it doesn't help enough to be worthy it. much. Careful design of experiments, selection of the right modeling methods, and scientific sampling will continue to have a significant impact. Relying on brute force alone, unaided by careful science, will give us brutish models.

------------------------------
David J Corliss, PhD
Analytics Architect / Predictive Analytics
Ford Motor Company

Original Message:
Sent: 12-01-2017 10:00
From: John Major
Subject: "Big data is dead. Data is just data."

This essay by Alexander Thamm provides a perspective I think most statisticians can align with:

https://www.linkedin.com/pulse/big-data-dead-just-regardless-quantity-structure-speed-thamm

 

 

John A. Major, ASA, MAAA

Director of Actuarial Research, Guy Carpenter Analytics

1166 Avenue of the Americas

New York, NY 10036

860-839-9148 (cell)

https://scholar.google.com/citations?user=9b7B9C4AAAAJ&hl=en

 

---

This message is intended only for the use of the addressee and may contain
information that is PRIVILEGED AND CONFIDENTIAL.

If you are not the intended recipient, you are hereby notified that any
dissemination of this communication is strictly prohibited. If you have
received this communication in error, please erase all copies of the message
and its attachments and notify the sender immediately. Thank you.

Original Message------

Is it possible that "Big Data" was replaced with "Machine Learning" and "Data Science" because those are the types of skills needed to analyze "Big Data"? 

For anyone that thinks data is just data, try working at Ford on autonomous cars or at U of Michigan health system taking signals from medical equipment. Both require taking live data from multiple streams of sensor arrays or arrays of sensors, and determining, "Is this data good?, If so, then what?" Get it right, no one knows or cares and things get better. Get it wrong, people die. 

Having worked with this type of data, it's not uncommon for each sensor to send 100Hz to 1,000Hz of signal data times X number of sensors times Y hours.... I think that gets pretty big. Thankfully, there is Machine Learning.

------------------------------
Andrew Ekstrom

Statistician, Chemist, HPC Abuser;-)
------------------------------

Original Message------

Is it possible that "Big Data" was replaced with "Machine Learning" and "Data Science" because those are the types of skills needed to analyze "Big Data"? 

For anyone that thinks data is just data, try working at Ford on autonomous cars or at U of Michigan health system taking signals from medical equipment. Both require taking live data from multiple streams of sensor arrays or arrays of sensors, and determining, "Is this data good?, If so, then what?" Get it right, no one knows or cares and things get better. Get it wrong, people die. 

Having worked with this type of data, it's not uncommon for each sensor to send 100Hz to 1,000Hz of signal data times X number of sensors times Y hours.... I think that gets pretty big. Thankfully, there is Machine Learning.

------------------------------
Andrew Ekstrom

Statistician, Chemist, HPC Abuser;-)
------------------------------

Well for the sake of the weekend, there is an old joke among astrophysicists and the rate of paper publication on dark matter, string theory or whatever topic you want to mock. It can easily be applied to Big Data. And it includes - as most jokes do - an important truth:

"Imagine all new data created in all connected networks (aka internet) as a growing pile of paper. The velocity of its growth has already exceeded c - or at least will do so soon enough.

Why is this not a contradiction to the theory of relativity?

- Simple: The information added with each new dataset is zero."

Best regards and a creative (or lazy - whicherver you prefer) weekend,
Christian


PS.: The challenge in 'Big Data' is often to pick the information from a huge pile of mostly redundant or unimportant stuff and not to fall for the noise. And yes - 'machine learning' helps a great deal with that. However, when I was still a student, we simply called it 'numerics' or 'applied mathematics' ;-).

------------------------------
Christian Graf
Dipl.-Math.
Qualitaetssicherung & Statistik

"To call in the statistician after the experiment is done may be no more than asking him to perform a post-mortem examination: he may be able to say what the experiment died of."

Ronald Fisher in 'Presidential Address by Professor R. A. Fisher, Sc.D., F.R.S. Sankhyā: The Indian Journal of Statistics (1933-1960), Vol. 4, No. 1 (1938), pp. 14-17'
------------------------------

Original Message:
Sent: 12-08-2017 09:35
From: Kelly Zou
Subject: "Big data is dead. Data is just data."

Some cartoon depictions of "big data" for some fun weekend reading!

 

http://www.ibmbigdatahub.com/infographic/four-vs-big-data

 

http://www.ibmbigdatahub.com/infographic/extracting-business-value-4-vs-big-data

 

 

 

Original Message------

Is it possible that "Big Data" was replaced with "Machine Learning" and "Data Science" because those are the types of skills needed to analyze "Big Data"?

For anyone that thinks data is just data, try working at Ford on autonomous cars or at U of Michigan health system taking signals from medical equipment. Both require taking live data from multiple streams of sensor arrays or arrays of sensors, and determining, "Is this data good?, If so, then what?" Get it right, no one knows or cares and things get better. Get it wrong, people die.

Having worked with this type of data, it's not uncommon for each sensor to send 100Hz to 1,000Hz of signal data times X number of sensors times Y hours.... I think that gets pretty big. Thankfully, there is Machine Learning.

------------------------------
Andrew Ekstrom

Statistician, Chemist, HPC Abuser;-)
------------------------------

What would they say about a method that does not even bother to gather new data but rather resamples data you already have? How redundant would that method be? 

If given the option of resampling the old or gathering the new, which is better?

------------------------------
Andrew Ekstrom

Statistician, Chemist, HPC Abuser;-)
------------------------------

Original Message:
Sent: 12-08-2017 10:31
From: Christian Graf
Subject: "Big data is dead. Data is just data."

Well for the sake of the weekend, there is an old joke among astrophysicists and the rate of paper publication on dark matter, string theory or whatever topic you want to mock. It can easily be applied to Big Data. And it includes - as most jokes do - an important truth:

"Imagine all new data created in all connected networks (aka internet) as a growing pile of paper. The velocity of its growth has already exceeded c - or at least will do so soon enough.

Why is this not a contradiction to the theory of relativity?

- Simple: The information added with each new dataset is zero."

Best regards and a creative (or lazy - whicherver you prefer) weekend,
Christian


PS.: The challenge in 'Big Data' is often to pick the information from a huge pile of mostly redundant or unimportant stuff and not to fall for the noise. And yes - 'machine learning' helps a great deal with that. However, when I was still a student, we simply called it 'numerics' or 'applied mathematics' ;-).

------------------------------
Christian Graf
Dipl.-Math.
Qualitaetssicherung & Statistik

"To call in the statistician after the experiment is done may be no more than asking him to perform a post-mortem examination: he may be able to say what the experiment died of."

Ronald Fisher in 'Presidential Address by Professor R. A. Fisher, Sc.D., F.R.S. Sankhyā: The Indian Journal of Statistics (1933-1960), Vol. 4, No. 1 (1938), pp. 14-17'

Original Message:
Sent: 12-08-2017 09:35
From: Kelly Zou
Subject: "Big data is dead. Data is just data."

Some cartoon depictions of "big data" for some fun weekend reading!

 

http://www.ibmbigdatahub.com/infographic/four-vs-big-data

 

http://www.ibmbigdatahub.com/infographic/extracting-business-value-4-vs-big-data

 

 

 

Original Message------

Is it possible that "Big Data" was replaced with "Machine Learning" and "Data Science" because those are the types of skills needed to analyze "Big Data"?

For anyone that thinks data is just data, try working at Ford on autonomous cars or at U of Michigan health system taking signals from medical equipment. Both require taking live data from multiple streams of sensor arrays or arrays of sensors, and determining, "Is this data good?, If so, then what?" Get it right, no one knows or cares and things get better. Get it wrong, people die.

Having worked with this type of data, it's not uncommon for each sensor to send 100Hz to 1,000Hz of signal data times X number of sensors times Y hours.... I think that gets pretty big. Thankfully, there is Machine Learning.

------------------------------
Andrew Ekstrom

Statistician, Chemist, HPC Abuser;-)
------------------------------

Original Message------

Is it possible that "Big Data" was replaced with "Machine Learning" and "Data Science" because those are the types of skills needed to analyze "Big Data"? 

For anyone that thinks data is just data, try working at Ford on autonomous cars or at U of Michigan health system taking signals from medical equipment. Both require taking live data from multiple streams of sensor arrays or arrays of sensors, and determining, "Is this data good?, If so, then what?" Get it right, no one knows or cares and things get better. Get it wrong, people die. 

Having worked with this type of data, it's not uncommon for each sensor to send 100Hz to 1,000Hz of signal data times X number of sensors times Y hours.... I think that gets pretty big. Thankfully, there is Machine Learning.

------------------------------
Andrew Ekstrom

Statistician, Chemist, HPC Abuser;-)
------------------------------

It has been fun to read this material and three thoughts occurred to me that may be helpful:
1. John Tukey in the early 60's defined big data as "a data set too big to fit on one device". The idea was that when you
needed multiple devices you needed much more complicated procedures to analyze them. I think his definition still holds. Of course, when Tukey waas writing he was talking about tape drives and now they are peta-byte devices  (and more), but the same idea is useful.
2. What is statistics if not a set of procedures to tell a lot about a large data-set from smaller samples.Since the precision of our measures goes as the square root of the sample we have, there is a diminishing return.
3. And what about continuous streams of data? Yes such streams, if digitized, contain immense numbers of data points, but such a digitization only approximates the continuous nature of the stream. Why not treat the stream as a single continuous function and use the tools of functi0nal data analysis pioneered by Jim Ramsay and his colleagues ?

Just my 2¢ on a Monday morning.

------------------------------
Howard Wainer
Extinguished Research Scientist
------------------------------

Original Message:
Sent: 12-08-2017 10:21
From: Kelly Zou
Subject: "Big data is dead. Data is just data."

Dear All,

 

Although "Big Data" may have various V's and interpretations, the term "Real World Evidence" has been defined in the U.S. 21^st Century Cures Act.

 

Please see my recent interview in AmStat News and a podcast with the ASA BIOP Section.

 

http://magazine.amstat.org/blog/2017/10/01/kelly-h-zou-on-real-world-evidence   

 

http://community.amstat.org/biop/podcast  

 

Happy Holidays!

 

Kelly

Chair, ASA SPAIG Committee

Chair-Elect, ASA HPSS Section

Original Message------

Is it possible that "Big Data" was replaced with "Machine Learning" and "Data Science" because those are the types of skills needed to analyze "Big Data"?

For anyone that thinks data is just data, try working at Ford on autonomous cars or at U of Michigan health system taking signals from medical equipment. Both require taking live data from multiple streams of sensor arrays or arrays of sensors, and determining, "Is this data good?, If so, then what?" Get it right, no one knows or cares and things get better. Get it wrong, people die.

Having worked with this type of data, it's not uncommon for each sensor to send 100Hz to 1,000Hz of signal data times X number of sensors times Y hours.... I think that gets pretty big. Thankfully, there is Machine Learning.

------------------------------
Andrew Ekstrom

Statistician, Chemist, HPC Abuser;-)
------------------------------