When is a statistic irrelevant?

[Pumpkin]

How small a sample can you get away with for a (meaningful) statistical analysis?

I've just noticed on the latest advert for Gillette's Fusion, touting it's benefits for sensitive skin that "85% of 68 men agree" with the statement "Great performance even on sensitive skin".

Surely a sample group of 68 is not representative? Even taking 6,900,000,000 as a conservative estimate of global population, halve that for males, knock 75% off that for those who don't shave (for various reasons) and you're left with 862million men who shave. Interestingly, Gillette quote the figure 600million...so let's go with their number. That makes those 68 guys about 0.0000113% of those shavers.

Conversely, there are 30,515 members here (Wow! When did B&B get so big?)...that makes us over 450 times more numerous!

P.S. Does anyone else think the footage of Roger Federer is all done with CGI?

 

[ouch]

Sounds like 85% more BS.

 

[Doc4]

I'd rather take my advice from 30 Helens ...

[YOUTUBE]http://www.youtube.com/watch?v=BDTZcj8Xink[/YOUTUBE]

 

[_JP_]

Laugh at the stat and move on down the shaving isle.

 

[i_shaved_something]

I wonder if there's 68 men on Gillette's Board of Directors....and if 15% of those are also on the Board of Directors at Schick who got confused.

 

[gollum83]

Well I guess it's marginally better than four in five dentists recommending Gillette Fusion.

 

[cheerwind]

Hehe.

To use a sample size of 68, I can guarantee you that they're using a ±10% margin of error and 90% confidence interval.

Yes, amazingly, 68 men is enough to represent billions of men, but again, ±10% margin of error and 90% confidence interval, which isn't exactly going out of their way in the name of accuracy. Basically, they went as loose as they could get away with (probably under the rules of marketing standards org. they are part of).

To get a ±3% margin of error and 95% confidence interval, they'd need a little over 1000 men, which would be the same as used by the Gallop poles you always hear about on the news.

 

[Stubblefield]

How small a sample can you get away with for a (meaningful) statistical analysis?

1024 or therabouts. But I'm only 98% confident.

 

[raisindot]

As a semi-professional researched and the author of a recent study on 401(k) plan sponsors, I can tell you that the number of respondents is less important than the methodology used to survey them.

For example, Gillette has testing labs in many cities. I've been to several of these sessions. You get paid for your time and try out stuff. Volunteering for one of these and knowing you're getting paid pushes you in the direction of favoring the product you're testing.

Also, the way subjects were tested may bias the sample. For example, they may have been told to use Kyle's prep before they shaved, which would have biased the results, since nearly any razor is going to shave better with Kyle's prep. Or the questions might have been written in a way to slant the results.

Take it with a grain of salt.

Jeff in Boston

 

[Crixus]

which would be the same as used by the Gallop poles you always hear about on the news.

As opposed to how the Warsaw ones use them--statistically the same, but actually quite different.

 

[Pumpkin]

Take it with a grain of salt

I did, whilst heading to the bathroom for a DFS with a Slim loaded with a Feather and some MWF...felt pretty good on my skin!

 

[dpm802]

For example, Gillette has testing labs in many cities. I've been to several of these sessions. You get paid for your time and try out stuff. Volunteering for one of these and knowing you're getting paid pushes you in the direction of favoring the product you're testing.

I once took part in a razor-testing research project.

I was given three disposable razors in beta-test stage, told to take them home and use them, and a researcher would call me in about 2 weeks to record my opinions. I was going to paid $10 or so for my part in the project.

The first razor was horrible. It tugged at my skin, didn't cut very well, and left me with rashes and irritations. I didn't like it. The second razor was no better. I was so disgusted by the first two, I didn't even try the third one.

When the researcher called to hear my opinions, I told him, honestly, that the products sucked. I didn't like them, I wouldn't buy them, I wouldn't recommend them to a friend, I wouldn't even shave with them if they gave them to me for free.

Guess what? I did not get paid for my participation in the project.

 

[DSB]

I am 87% positive that statistics are irrelevant 64% of the time. Of course, I'm wrong 72% of the time. On the other hand, I'm right 91% of the time during leap years.

 

[mmack66]

“There are three kinds of lies: lies, damned lies, and statistics.”

 

[Obsessed]

I just hope I'm not going to the fifth dentist.

 

[instpasr]

aren't 87% of all statistics made up.

 

[wdwrx]

Make it big, repeat it often.

I find it rather amusing that in Gillette's new commercial that they are tacitly retracting their entire "lift and cut" dogma of the past, and now passing the new version off as an improvement as it no longer "lifts and cuts".

 

[ModusPwnins]

As a semi-professional researched and the author of a recent study on 401(k) plan sponsors, I can tell you that the number of respondents is less important than the methodology used to survey them.

This. A truly random sample will usually give you representative results even with a sample size as low as thirty!

 

[FOGEY]

87% of 99% of all statistics are 100% wrong!




When is a statistic irrelevant?

When it is not in your favor!

 

[wimbouman]

How small a sample can you get away with for a (meaningful) statistical analysis?

I've just noticed on the latest advert for Gillette's Fusion, touting it's benefits for sensitive skin that "85% of 68 men agree" with the statement "Great performance even on sensitive skin".

Surely a sample group of 68 is not representative? Even taking 6,900,000,000 as a conservative estimate of global population, halve that for males, knock 75% off that for those who don't shave (for various reasons) and you're left with 862million men who shave. Interestingly, Gillette quote the figure 600million...so let's go with their number. That makes those 68 guys about 0.0000113% of those shavers.

Conversely, there are 30,515 members here (Wow! When did B&B get so big?)...that makes us over 450 times more numerous!

P.S. Does anyone else think the footage of Roger Federer is all done with CGI?

Even more interesting, they do not claim these 68 persons to be a sample group.

That probably means it is just 68 testers.

And their claim is just literally as stated: 85% of those (my guess is 57 people) are happy with that new Gillette product.

No statistics whatsoever.