Agnostic.com

3 0

Statistics do not lie. Statistical tests are just that, tests aimed at interpreting patterns and trends. They are mathematical applications to information. Statistical outcomes are not sentient, they do not have opinions.

Statistical tests are only as good as the questions they are designed with, and the sample size of the data collected. A statistical test laced with biased questions with net a biased outcome. Likewise, too small of a sample will net an inclusive or insufficiently supportable outcome.

Is the researcher doesn't understand the nature of the statistical test they have selected to apply to their research (all too often the event in political topics) they arrive at an erroneous result.

Likewise, if the interpretor doesnt understand what the statical test is designed to test, their interpretations become misleading and incorrect at best, fanciful at worst. The best statistical test are designed with neutrally asked questions, those that willfully are designed to minimize bias of any kind.

t1nick 8 Nov 7
Share

Enjoy being online again!

Welcome to the community of good people who base their values on evidence and appreciate civil discourse - the social network you will enjoy.

Create your free account

3 comments

Feel free to reply to any comment by clicking the "Reply" button.

0

Sounds like a plea in support of Eugenics. Its originator, Francis Dalton, also laid the foundations for what became statistics.

[en.m.wikipedia.org]

Hmmm, interesting observation. I would for all practical purposes be the last person to support eugenics. Again, statistics are only as good as the questions asked. Biased questions yield biased outcomes.

@t1nick Statistics are at their most useful when no questions are involved. In other words, when past outcomes are observed rather than future outcomes asked.

@Petter

The basis for any research begins with a question. You cannot do research without beginning with a queztion, no matter what you are researching.

@t1nick The question should preferably begin "what happened when ..."
Note the past tense.
Then it can be tested by "What happens if ..."

@Petter

Questions are tailored according to the intended research. There is no such thing as one question that fits all resesrch

@t1nick Like wife beating, for example. 🤣🤣

@Petter

When I teach my students how to generate a hypothesis, I use an IF >THEN>WHEN structure. The IF statement sets up the original condition (Control variable). The THEN statement introduces the variable to be changed (Independent variable), and the WHEN statement re-states the variable change, then makes the prediction (actual hypothesis) and represents the Dependent variable.

@t1nick In my coding days, <when> statement meant every time a global variable changed, it was to be retested against a condition, and if positive a new action selected.
It was very useful in reducing the size of coding, whilst speeding up the reaction. For example, in a space invaders game.
WHEN Missile=Enemy : Explode.

@Petter

Its been awhile since I wrote any code. But I remember

@t1nick A little over 30 years in my case. Except simple HTML code, but that doesn't count.
My children, though, are both serious coders, and my 12 years old grandson is developing some nifty stuff.

1

Ooh, I could go to town with this one ...

  1. When did you stop beating your wife?
  2. How big was your last donation to the Republican Party?
  3. What have you done in the last 10 years to prevent a repetition of Sandy Hook?
  4. Why do you think that there are more people of colour in our jails?
  5. What does God want you to do?
  6. Why did the government pull the Covid-19 hoax?

Your point? A statistical test is only as good as the questions asked. A good statistical analysis should be worded as unbiased and as neutral as possible, yet aimed as accurately as possible towards the target isdue.

@t1nick My point was humour. I threw together a list of heavily biased questions.

@anglophone

Ahhhh! Lol.

1

Statistics could be manipulated by a toddler to get whatever outcome you wish...notoriously so!

No statistics cannot. Interpretation by unscrupulous people and uninformed, gullible consumers can be. Statistics cannot. If your statistics are manipulated, then ther outcome is worthless. Statistics say what statistics are designed to say, nothing more, nothing less. A poor test says more about the designer than it does about the test.

@t1nick Agreed. My own observation is that some statistics are foisted on an unsuspecting public by people who should know better. My favourite is "Trans-Monaco Airlines has never had a flying accident [but we don't have any planes!].".

@anglophone

You illustrate an interesting point. What responsibility do average American citizens, or the citizens of any First World country, have to educate themselves in assessing credibility of any source. What this last 5 years surrounding our Trump experience has illuminated is that, at least in the US, is that our citizenry has become intellectually lazy (not really new news).

But have taken this intellectual laziness beyond what would be expectef of a nation of our resources and education access. The eagerness to suspend reasonable certainty and facts for unreasonable conspiracies and unsubstantiated claims. It befuddled the mind that we have slid so far down this rabbit hole.

You can include a link to this post in your posts and comments by including the text q:550580
Agnostic does not evaluate or guarantee the accuracy of any content. Read full disclaimer.