# How does regression analysis help one understand how the typical value of the dependent variable change?

regression analysis helps one understand how the typical value of the dependent variable changes... -- http://en.wikipedia.org/wiki/Regression_analysis

What does this mean? What "typical value" is it talking about? What kind of change is it looking for? Positive or negative or what? Change in what? Percentage? I'm confused.

Lessons learnt:
* wikipedia documents a ton of info, most of which are unreadable and unusable
* most (80%) ppl, while im sure good meaning, just dont know how to give good answers
* and only human... have so many flaws, and thereby closes good questions

-
Thank you, this is a better question, & one that I think is answerable. –  gung Sep 7 '12 at 3:12
the original question was answerable -- stats.stackexchange.com/questions/35824/… -- someone telling me to split the question into many got me into problems –  kittensatplay Oct 7 '12 at 6:28

Say you wanted to guess someone's weight, but you only know their height. How could you do this?

If you had access to measurements of the the height and weight of many people you could draw up a graph with height and weight on the axes and put dots at the points of each pair of measurements. What you will see is a general trend for taller people to be heavier, but there is a lot of scatter around this trend.

Regression is the process of using some technique to estimate the typical relationship between the height and weight. So, after doing the regression analysis you might be able to say

Typically, someone who is 1.8 metres tall weights about 80Kg, but someone 2
metres tall weighs around 100Kg.


Of course plenty of people who are 1.8 metres tall weigh more or less than 80Kg, but your analysis might tell you that in the absence of any other information your best guess for their weight is 80Kg. The regression analysis tells you how this typical value for weight changes with height i.e. taller people are typically heavier.

Regression is not limited to one variable predicting another, for instance what if we also knew the biological sex of each person? After doing some regression analysis we might be able to say

Males that are 1.8 metres tall typically weigh around 85Kg but Females
weigh around 75Kg.


Note that I just made up all these numbers. The actually numbers would depend on doing some regression analysis on a real data set (which I haven't done for this quick explanation).

-
this is SOOOOO much clearer. can you PLEASE edit up wikipedia because it's SO BAD. it will help the world and society at large. –  kittensatplay Sep 7 '12 at 3:34
It's tricky. If I was to re-write the Wiki article in this style it would be 10 times longer. The maths and science articles on wiki do tend to be written to a higher level than others and for the most part this is fine because they function as a quick reference for the kind of people who read and edit those pages. If I wanted to learn about post-modernist feminist literary criticism from scratch I would have to accept that wiki won't give me all the details, but might function as a reference for where to find those details. –  Bogdanovist Sep 7 '12 at 3:41
There is the alternative of the simple english version of wiki, which is supposed to be written at the lowest level. I just check the regression page simple.wikipedia.org/wiki/Regression_analysis and it is pretty disappointing. That might be worth someone improving in the way you are arguing for. –  Bogdanovist Sep 7 '12 at 3:46
oh god, im almost done analyzing this, and this is SOOO yummy. i could make this answer more concise from a beginner's viewpoint but the key thing is you made relevant what is needed that the wiki article doesnt. learning from a typical wiki article would be hard for anyone entry-level. for anyone advanced, well, they probably should've learned it this way from the beginning -- instead of the way most "typical" teachers/textbooks/webpages would teach it. it really really should replace the lead section of the article, but wikipedia wont allow it. the simple version is 10x worse... honestly –  kittensatplay Sep 7 '12 at 3:52
@kittensatplay: I think it's better for wikipedia to be a good reference rather than a tutorial. Textbooks make better tutorials, and they can even accomodate the student better because you can have a different textbook for each approach and learning style. Have you considered getting a statistics textbook? –  Neil G Oct 7 '12 at 9:53