10

# Anecdotes are not "piles of statistics".

Mon 26 Jan, 2015 08:43 pm
This is from another thread, where anecdotes were presented as statistics. This is a common logical fallacy that leads to bad assumptions including exaggerating the risk of crime or epidemic.

We have a saying in engineering "data is not the plural of anecdote". There is a mathematical reason for this.

Valid statistics present a whole picture. When you consider statistics you take into account the cases that contradict your preconceived notions as well as the ones that support it.

Anecdotes are not objective. No one ever selects an anecdote that challenges their own assumptions. Anecdotes are offered as one case which may or may not represent the whole truth, the contradictory cases aren't considered.

Just as important, statistics can be questioned. They are measured by an objective process that can be examined and questioned. This is an important part of any honest search for truth.

Any honest objective debate is based on statistics, not anecdotes.
• Topic Stats
• Top Replies
Type: Discussion • Score: 10 • Views: 2,778 • Replies: 35

CalamityJane

2
Mon 26 Jan, 2015 08:48 pm
@maxdancona,
For every statistic you bring, I'll get you one saying the opposite. Statistics are not something you should bank on.

Anecdotes, on the other hand, are stories of interest that may or may not be true. Has nothing to do with statistics.
maxdancona

1
Mon 26 Jan, 2015 09:08 pm
@CalamityJane,
Quote:
For every statistic you bring, I'll get you one saying the opposite.

When I say statistics, the implication is that they are collected with a scientific process. The advantage is that they can be questioned... some statistics are valid and some statistics are invalid. A scientifically done study will yield statistics with the ability to give you an objective view.

Ticomaya

4
Mon 26 Jan, 2015 09:13 pm
@maxdancona,
No, but statistics are "piles of anecdotes." Take a bunch of anecdotes, and you got statistics.

You may try a prove your case by citing to statistics, but I can point to my anecdote to prove that -- at least in that one instance -- your statistics were wrong.
Lustig Andrei

2
Mon 26 Jan, 2015 09:14 pm
Everyone knows that 78.3 % of all statistics are made up out of whole cloth.
0 Replies

cicerone imposter

1
Mon 26 Jan, 2015 09:19 pm
@maxdancona,
Yea. It's about the bell curve.
0 Replies

2
Mon 26 Jan, 2015 09:25 pm
@maxdancona,
Arguably you are comparing two different research methods: qualitative and quantitative.

But absolutely resent, in my field, a single anecdote being use as an argument against a bunch of reproducible data. Am over librarians who say they are into 'evidence based librarianship' then ignore data based on what some friend told them.
maxdancona

1
Mon 26 Jan, 2015 10:00 pm
@Ticomaya,
Tico, valid statistics aren't about "one instance", they are about facts as a whole. There are anecdotes about people who have died as a result of them wearing a seat belt. This doesn't change the fact that wearing a seat belt makes you (or people in general) much less likely to die in an automobile accident.

I think you are missing my point in general.

Any bunch of anecdotes I give you aren't going to reflect any objective facts... they are going to reflect my prejudice. That is what anecdotes are for. I select them to support my preconceived notions.

Valid statistics, on the other hand, are gathered through an open, objective process that may or may not contradict my personal bias.

I have seen valid statistics contradict the pre-conceived opinions of the people who gathered them (in fact, I have seem them contradict my own pre-concieved opinions).

Have you ever seen someone give an anecdote that has contradicted their personal bias?
maxdancona

1
Mon 26 Jan, 2015 10:03 pm
Quote:
But absolutely resent, in my field, a single anecdote being use as an argument against a bunch of reproducible data. Am over librarians who say they are into 'evidence based librarianship' then ignore data based on what some friend told them.

Refusing to accept reproducible data on account of a single anecdote is logically invalid. I don't think this has anything to do with qualitative vs. quantitative.

0 Replies

Ticomaya

2
Mon 26 Jan, 2015 10:23 pm
@maxdancona,
You might tell me, as one other poster at this forum has done in the past, that -- statistically speaking -- a child is more likely to die in a car crash than by a pit bull attack; regardless, I'm not going to let my young child play with a pit bull. Why, because -- anecdotally speaking -- I have personally seen the mutilation caused by a pit bull to a 10 year old boy. That's an anecdote I bring to the table. Now, does that anecdote reflect my bias, or does my bias reflect my anecdote?
Kolyo

1
Mon 26 Jan, 2015 10:24 pm
@CalamityJane,
CalamityJane wrote:

For every statistic you bring, I'll get you one saying the opposite. Statistics are not something you should bank on.

Statistical claims have a degree of subjectivity to them, because the definitions they are based on are subjective. My claim that America's poverty rate is above 99% is valid if I define "poverty" to mean "having an income of less than \$1m per year". Your counter-claim that the poverty rate is below 1% is equally valid if you define "poverty" to mean "having to go for a fortnight without eating".

Using a statistic that is based on terms that are defined differently than in the standard way is just one way to lie with statistics.
0 Replies

2
Mon 26 Jan, 2015 10:30 pm
@maxdancona,
Quote:
Any bunch of anecdotes I give you aren't going to reflect any objective facts... they are going to reflect my prejudice. That is what anecdotes are for. I select them to support my preconceived notions.

This bit is what I meant about qualitative vs quantitative. A bunch of anecdotes can indeed by classified as data - if you accept that a survey, for example, is a way of gathering anecdotal evidence. Indeed statistics themselves are used to analyse and validate this 'anecdotal' evidence.
0 Replies

maxdancona

1
Mon 26 Jan, 2015 10:33 pm
@Ticomaya,
Tico, this example has nothing to do with bias. This is an example of a logical fallacy. There is at least one obvious mathematical error in the argument being offered by this poster.

Valid statistics answer a question in an objective, scientific way. You still have to ask the correct question.

Anecdotes don't answer any question objectively. Of course they will get an answer persons bias happens to be correct. But to know an answer is correct (in an objective sense) you need an scientific, testable way to get the answer that isn't affected by bias.

Kolyo

3
Mon 26 Jan, 2015 10:33 pm
@Ticomaya,
Ticomaya wrote:

You might tell me, as one other poster at this forum has done in the past, that -- statistically speaking -- a child is more likely to die in a car crash than by a pit bull attack; regardless, I'm not going to let my young child play with a pit bull.

Another way to lie with statistics is to present irrelevant statistical facts. It is true that a given child from the total population of children in the USA is more likely to die in a car crash than from a pitbull attack, but that's irrelevant to someone deciding whether to let his children play with pitbulls. What's relevant is that a given child from the total population of children who play with pitbulls is more likely from a pitbull attack than in a car crash.
maxdancona

1
Mon 26 Jan, 2015 10:36 pm
@Kolyo,
Kolyo,

Do you believe that there is a mathematically valid objective way to answer questions about things like risk? Note that I am using the term "valid statistics" by which I mean statistics gathered by a scientific objective process. Do you agree with this term?
cicerone imposter

1
Mon 26 Jan, 2015 10:41 pm
@maxdancona,
maxdancona

1
Mon 26 Jan, 2015 10:47 pm
@cicerone imposter,
I don't think so Cicerone. There are many questions that you can ask objectively... and at times you will receive answers that contradict your preconceived notions. Some important discoveries in science were made this way.

I have asked objective questions and received answers that forced me to change my mind, haven't you?

If all answers were biased, why would anyone ever change their mind?
Kolyo

2
Mon 26 Jan, 2015 10:54 pm
@maxdancona,
Yes. Both the statements I cited in my previous post are mathematically valid statements about risk.* In the first case, I noted that the "prior" probability of a kid dying from an auto accident was greater than the prior probability of the kid dying from pitbull attack. In the second statement, I noted that the "conditional" probability of a kid dying from pitbull bite, given he plays with pitbulls, was greater than the conditional probability of the kid dying from an auto accident, given he plays with pitbulls.

Both statements can be technically true without there being a contradiction. But if you use the first statement to make a decision on whether your son should play with pitbulls, you're using an irrelevant, misleading statistical claim. And if you know of three fathers who let their sons play with pitbulls, and if one those sons is known to died in the act, you would do well to consider that anecdotal evidence.

*--I would guess. I'm not sure about the second one.
cicerone imposter

1
Mon 26 Jan, 2015 11:19 pm
@maxdancona,
True and often; we try our best.
0 Replies

maxdancona

1
Tue 27 Jan, 2015 06:08 am
@Kolyo,
Quote:
Both statements can be technically true without there being a contradiction. But if you use the first statement to make a decision on whether your son should play with pitbulls, you're using an irrelevant, misleading statistical claim. And if you know of three fathers who let their sons play with pitbulls, and if one those sons is known to died in the act, you would do well to consider that anecdotal evidence.

Hmmmm.... I fail to see the value of anecdotal evidence even in this case.

Sure... in this single example, the anecdotal evidence happens to yield the logically correct answer. However, many times the anecdotal evidence will lead you astray.. sometimes dangerously so (as it did in the past with cigarette smoking or seat belt usage).

To know whether anecdotal evidence is valid or not, you still need real valid (research based) statistics anyway.

I can give you anecdotes of kids dying from playing baseball. How would you suggest I decide if baseball is more important then playing with pitbulls?

### Related Topics

New Propulsion, the "EM Drive" - Question by TomTomBinks
The Science Thread - Discussion by Wilso
Why do people deny evolution? - Question by JimmyJ
Are we alone in the universe? - Discussion by Jpsy
Fake Science Journals - Discussion by rosborne979
Controvertial "Proof" of Multiverse! - Discussion by littlek

1. Forums
2. » Anecdotes are not "piles of statistics".