You are invited to Log in or Register a free Frihost Account!

# You can proof anything with statistics

Coursera Statistics One" by AJ Cann.

In the recent ‘bild der wissenschaft' magazine has been an article about self-driving cars and the question is asked whether that one crash caused by a Tesla car will severely hinder the adoption of self-driving cars.
The short answer is no, such rare cases of malfunctions will help to improve technology over time, and overall self-driving cars are safer than cars driven by humans. They show a pie chart indicating that 93 % of car accidents are caused by humans while 1 % are caused by technical failures. So far, so good. This kind of suggests that we really should not hesitate to get started with self-driving cars. Whether this statistic will be true in the future is something nobody can answer.

Lifehacker published an interesting article today: "How to Lie to Yourself and Others With Statistics".

You probably know that saying: never trust any statistic unless you have created faked it yourself.
Or that one:
>> There are three kinds of lies: lies, damned lies, and statistics. << Benjamin Disraeli, British politician (1804-1881)

I love statistics. Particularly when one puts them all together in an Excel sheet and then looks for a pattern.

I agree statistics can be misrepresented. Or manipulated for a given outcome. Thank goodness if it is used for science, it gets to be peer reviewed, and unmasked.
deanhills on Wed Oct 26, 2016 10:59 pm
You might be surprised just how many scientific studies are published with faulty statistics. Either the wrong types of tests are used, assumptions of the tests are violated, sample sizes are too small, or are interpreted incorrectly.
Of course, part of reading a published study should include an assessment of the methods used and their validity... but we're not always well equipped as researchers to properly understand the statistics being used. We usually have some grasp, but statistics is a full study in and of itself, and the devil can be in the details in a lot of cases

Responsible reporting and use of statistics should be transparent, and those reporting should make efforts to ensure that the method of reporting isn't biasing interpretation... but, that doesn't really happen
Ankhanu on Thu Oct 27, 2016 4:01 pm
Statistics can so easily be manipulated to show desired results. Just like election polls.
standready on Thu Oct 27, 2016 4:44 pm