This comment was never valid. At least not in the direct meaning of the words. It is just a sarcastic statement. It is very dishonorable to present statistical "facts" to convince others that are not aware of the context in what the statistic was derived.

Results of statistical analyses (be it just averages or percentages!) can ONLY be understood in the whole context of the question. But people usually think they unterstand everything if they are confronted with simple, easy-looking numbers (e.g. the rate of unimployment is 12%; the average CO2 emmission per person and year is 4600kg ... same with changes and correlations). Deliberatel using this misconception to influence and to steer people with the aim to achieve personal advantages is common... and a really despicable behaviour. This is still true.

This comment was never valid. At least not in the direct meaning of the words. It is just a sarcastic statement. It is very dishonorable to present statistical "facts" to convince others that are not aware of the context in what the statistic was derived.

Results of statistical analyses (be it just averages or percentages!) can ONLY be understood in the whole context of the question. But people usually think they unterstand everything if they are confronted with simple, easy-looking numbers (e.g. the rate of unimployment is 12%; the average CO2 emmission per person and year is 4600kg ... same with changes and correlations). Deliberatel using this misconception to influence and to steer people with the aim to achieve personal advantages is common... and a really despicable behaviour. This is still true.

I am 100% agree with Wilhelm. This is just sarcastic statement from Non statisticians who are not willing to understand statistics in real scence. It is science to predict averagely phenomena in the population. One has to be careful enough to use appropriate statistical tools before embarking on results.

Statistics are never a source for lies. They are neither the absolute truth. Correct statistical inference, and careful interpretation is always necessary. The most important thing about statistics is how they are handled. Recall Box's famous motto: "All models are false, but some of them are useful".

As other respondents have stated, it all depends on the correct use and careful interpretation of statistical results. Therefore, researchers with malign intents can indeed use statistics as a powerful tool for lying. See the amusing paper of Wainer (1984), "How to display data badly", which exposes some tricks to curb the truth. The article is a bit old, but still a nice reading.

Keep a critical mind, always keep the context in mind and double-check numbers as much as possible.

As far as I have understood, the word ''statistics' in the comment was used to mean 'raw data'. Indeed, in 1891 when this comment was made, hardly anything in the name of data analysis was in existence.

.

Hemanta, you're right : "statistics" here means "data", indeed !

But "half right" only : statistics, even in those early days, were not "raw data" (a funny notion, by the way ; do data grow spontaneously somewhere where statisticians just have to pick them ?) but agregates on various pre-defined categories ; this is where the "context" and the necessary common agreement (or at least full information about the defintions of the agregates/categories used by the other party) enters, as indicated by Jochen above.

(the many conflicting definitions of "unemployment" are nothing new ... so what is "the" unemployment rate ?).

interesting related link :

"Six ways to separate lies from statistics", Betsey Stevenson & Justin Wolfers

(in summary :

1- Focus on how robust a finding is, meaning that different ways of looking at the evidence point to the same conclusion.

2- Don’t confuse “statistically significant” with something actually mattering.

3- Be wary of scholars using high-powered statistical techniques as a bludgeon to silence critics who are not specialists.

4- Don’t fall into the trap of thinking about an empirical finding as “right” or “wrong.” At best, data provide an imperfect guide.

5- Don’t mistake correlation for causation.

6- Always ask “so what?”)

http://www.bloomberg.com/news/2013-05-01/six-ways-to-separate-lies-from-statistics.html

as far as sarcastic remarks about statistics are concerned, i always loved Sir Winston's “Statistics are like a drunk with a lampost: used more for support than illumination.” !

(note that, this time, the comment is not about statistics themselves, but about their usage)

.

Modern statistics is omnipresent in almost every branch of scientific investigation. So the comment is invalid for all those who are not liar.

The appropriate flip response to the flip statement is "Statistics don't lie but liars use statistics". Statistics are just numbers calculated using formulas. The truth or lie comes from the interpretation of statistics. The basic truth required for truthfully using a statistic is to understand, evaluate, and summarize how well the statistic works with your data in language that can be understood by the people to whom you are describing it. To me, the most important truth is to then use plain language in summarizing what that statistic means in the real world.

Whether or not Benjamin Disraeli or Mark Twain coined the expression, the root of the sarcasm remains valid. Many of the comments addressed the willingness of the speaker to distort to accomplish an agenda. There is the reverse side of statistics: the listener. Am I a gullible listener who believes conclusions drawn from a sampled set of statistics or am I a discerning analyst. While Twain may have been pulling the chain of the speaker, he was also pulling the chain of the listener (e.g. how well do I know the model implicit in statistical analysis, how well do I understand the proof requirements of the SUT, ...). When we listen, we also have the opportunity to lie about the validity of what we are listening to or to discern the accuracy of that speaker.

Dear Snashall,

Data Science has already started to become a new subject of study. Big Data Analysis too is coming up. These are computer dependent fields. Some people will create some software, and some others will use them. Most of the users would perhaps never bother to know whether the available software is actually applicable in their respective cases. In the process, mathematics would perhaps be a forgotten thing!

So whenever I remember this comment, I feel, perhaps Disraeli was right after all!

I think that there is an interesting paper on the proper and improper use of models was written in 1973:

Westfall, Richard S., Newton and the Fudge Factor, Science, Volume 179, Number 4075, 23 February 1973, pp 751-658.

It is a study not only about how the author uses analysis to obscure the issues but also in how the receiving audience (Royal Society) aids and abets the FUDGE of data by Newton. Even more so would be the emergence of the calculus with Newton credited as the first and primary discoverer in his control of the publications by the Royal Society. My point is that although statistics is based on assumptions that can be validated (or at least justified) by analysis, it is up to the reader to discern whether or not this is real science. In Newton's case, he controlled the evidence flow and that made discernment difficult. Still, the reader must rise to the level of skeptical reading to provide a rejoinder to the works of others.

To remember the saying is important for any reader of scientific (or political) literature. You are right about this maxim!