I need Statistical data!

Dijos

Dijos

Joined
Nov 20, 2001
Messages
2,600
I have Stats 1, and we have a project due in about 2 weeks. It's supposed to be a simple linear regression or something fairly simple. I need raw data, and want to do something on knives (duh). Anyplace I can find a good amount of data? any suggestions on what I should look for/how I should put it together? Correlation between heat treat and edge chipping or something? any and all suggestions will be appreciated, as I hate Stats, and want to make it fun.--Joe
 
Try contacting CATRA, Spyderco, or Buck knives. Spyderco and Buck both do a lot of CATRA testing, goniometer testing, etc. They might be able to give you some data. If you do get data from a company, I'd suggest sending them a copy of your report as a thanks.
 
Gator, linear just means in the model parameters, not the actual functions.

Specifically :

f(x)=sum a_i * g_i(x)

is a linear model

so for example

1+a*x^10+c*Exp(-x)

is a linear model with paramters a and c, whereas

Sin(a*x)

isn't linear because it isn't of the form

<some parameter>*<somefunction>

Note the vast majority of nonlinear problems can be made linear with approximations.

As for statistics, these are trivial to collect. For example go into the for sale forum and take the last one hundred posts, find correlations between

-the price and the number of replies
-the price and the number of knives of that type sold
-the members # of posts and the price of the knives sold

Some of these are likely to turn up strong very simply correlations of the form a*x, some of them will show nothing outside of random correlation.

It is also trivial to gather such statistics in everyday life. The next time you are eating lunch say hello and smile at a random person who passes by then se if you can correlate :

-their age (estimate) to the number of returned smiles and hellos
-look at the influence of sex, do you get more returns from men or women (noting that an average is just a linear regression of a constant function).
-try different phrases, do the number of words/syllables correlate to the number of responces (hello,hi,nice day,good afternoon)

Use your imagination.

-Cliff
 
Hi Dijos,

If you want to get through your project easily listen to Cliff. He has played the academic game extensively and knows that easy to get data will give you just as good a grade as interesting data. I have two sons who routinely ignore this advice and are punished for their ambition by stress and sometimes bad grades (the project gets turned in late while they're trying to get that significant/interesting data). You might just compose a question to ask knife owners on this forum. Just try and make it a question with a couple numbers to correlate. Example: how many knives have you purchased in the last year. From their posting you can find how long they have belonged to Bladeforums and how many times they have posted. You can correlate knives-per-year to postings-per-week. Something like this will require you to do a little manual data collection, but you could do it in a couple days.

If you want to find some interesting data you should check on the Buck forum. I know that they posted some graphs about their CATRA testing of different alloys and blade grinds. I was really surprised at the time that they would share such proprietary information.

This is a link to the original thread, but it looks like the link to the graphs is broken:

http://www.bladeforums.com/forums/showthread.php?t=127499&highlight=CATRA
 
You must log in or register to reply here.
Back
Top