UnderGround Forums
 

AcademicGround >> Elementary Statistics


9/6/12 9:05 AM
Ignore | Quote | Vote Down | Vote Up
groundfighter2000
226 The total sum of your votes up and votes down Send Private Message Add Comment To Profile

Member Since: 1/1/01
Posts: 39543
 
So I'm taking an Intro to Stat class this semester.

Wondering if anyone has any recommendations for websites or review books or anything. Math was never my strong suit but I'd like to own this class.

It will be covering Descriptive Statistics, Probablity and Inferential Statistics. And the prof said we'll only be dealing with univariate stats.
9/30/12 9:02 AM
Ignore | Quote | Vote Down | Vote Up
Roly_Poly_Puppy
Send Private Message Add Comment To Profile

Member Since: 1/1/01
Posts: 1371
Beecher et. al. book College Algebra if you need a brush up in elementary algebra, which will be important if it is a non-calculus based stats class.

For good introductory texts that have a ton of worked examples and exercises for practice:

Bluman's Elementary Statistics: A Step by Step Approach.

Moore, McCabe, and Craig's Introduction to the Practice of Statistics.

The good thing about these textbooks are that they have feedback in the form of answer keys in the back. I hate math textbooks that don't have feedback, and there are a shitload of textbooks like that (though it becomes important towards the end of your degree to ween yourself off the feedback, as problems in the real world won't have Instructor's solutions, but in first year and second year, instant feedback is a key part of learning).

If you need more practice problems to grind through, try the Schaum's series. There is a ton of different books. I think the latest Statistics Schaum's guide (4th ed from memory) would probably be ok.

You may also have to do some programming. Not sure what language you will be using. If you don't use a language like R, MATLAB, or SAS, then you may end up just using Excel, which isn't really a language, but whatever. Salkind's Statistics for people (who think) they hate statistics covers excel. For R, just get the R Cookbook from the O'Reilly publishers. The intro books above cover parts of minitab, and some other languages.

If you are struggling with counting problems, which underlie most probability, try Niven's Mathematics of Choice: How to Count without Counting. If you want something even more basic for stats try The Cartoon Guide to Statistics (I'm not kidding, this is a good book). If you want something big picture & philosophical, try Hacking's An Introduction to Inductive Logic and Probability. When I say philosophical, don't think of it as not practical. The book is excellent, and has a huge section on using probability in real life moral and decision making.

Post-stats course, look into two books:

Good and Hardin's Common Errors in Statistics (and how to avoid them), and;

Saville and Wood's Statistical Methods: The Geometrical Approach. You need Linear Algebra for this though, which Beecher's book above covers a little. This book will give you an intuitive understanding of statistics, as it teaches stats through geometry (data becomes vectors, hypotheses in hypothesis testing become the directions of vectors).


9/30/12 9:10 AM
Ignore | Quote | Vote Down | Vote Up
Roly_Poly_Puppy
Send Private Message Add Comment To Profile

Edited: 09/30/12 9:15 AM
Member Since: 1/1/01
Posts: 1372
Oh, and another really basic book like the Cartoon guide might be How to lie with statistics by Huff.


This book is brilliant, and in more ways than one. Go read the Amazon reviews and you'll see plenty of people praising the book. No where will you find reference to the historian of science Robert Proctor's book Golden Holocaust, which outs several famous statisticians (or statistics popularizers like Huff), as shills for the tobacco industry. So Huff's book is good to learn from, as you are learning from a real life scumbag.

more on andrew gelman's blog: http://andrewgelman.com/2012/04/how-to-mislead-with-how-to-lie-with-statistics
9/30/12 1:33 PM
Ignore | Quote | Vote Down | Vote Up
groundfighter2000
226 The total sum of your votes up and votes down Send Private Message Add Comment To Profile

Member Since: 1/1/01
Posts: 39968
Hey Roly Poly, thanks for the suggestions I actually took the "Bluman's Elementary Statistics: A Step by Step Approach" out of the library.

The class isn't so bad. Our first exam is a week from wednesday so far we done a quick overview of descriptive statistics. The only thing that's tripped me up a bit cause i don't think the instructor explained it well was when and how to use Chebyshev's rule (vs. using the Empirical Rule).
12/10/12 5:53 PM
Ignore | Quote | Vote Down | Vote Up
groundfighter2000
226 The total sum of your votes up and votes down Send Private Message Add Comment To Profile

Member Since: 1/1/01
Posts: 41131
Question about confidence intervals and critical values. We have tables for the critical values for different confidence levels. But if a question asks for a confidence level not on the chart (the chart gives 80%, 90%, 95%, 98%, 99%) i.e. 94%, how do i find out the critical value?
12/16/12 2:47 PM
Ignore | Quote | Vote Down | Vote Up
PhinPhan
49 The total sum of your votes up and votes down Send Private Message Add Comment To Profile

Member Since: 2/6/06
Posts: 1758

I'm assuming that you are talking about critical values of z for a 2-sided test. In that case, you express your confidence level as a decimal, add 1 to it, and divide the result by 2. Then you consult a table of the cumulative standard normal distribution to get the z value. I know that sounds complicated, so let's look at an example.

For a confidence level of 95%, the decimal is 0.95.

(0.95 + 1)/2 = 1.95/2 = 0.975

The z value for 0.975 is 1.959963985, which rounds to 1.960. If we're on the same page here, then your chart should show 1.960 as the critical value for 95% confidence.

Here is a brief table of critical values for different confidence levels:

90%   1.645
91%   1.695
92%   1.751
93%  1.812
94%  1.881
95%  1.960
96%  2.054
97%  2.170
98%  2.326
99%  2.576
 
12/16/12 3:02 PM
Ignore | Quote | Vote Down | Vote Up
groundfighter2000
226 The total sum of your votes up and votes down Send Private Message Add Comment To Profile

Member Since: 1/1/01
Posts: 41138
PhinPhan - 

I'm assuming that you are talking about critical values of z for a 2-sided test. In that case, you express your confidence level as a decimal, add 1 to it, and divide the result by 2. Then you consult a table of the cumulative standard normal distribution to get the z value. I know that sounds complicated, so let's look at an example.

For a confidence level of 95%, the decimal is 0.95.

(0.95 + 1)/2 = 1.95/2 = 0.975

The z value for 0.975 is 1.959963985, which rounds to 1.960. If we're on the same page here, then your chart should show 1.960 as the critical value for 95% confidence.

Here is a brief table of critical values for different confidence levels:

90%   1.645
91%   1.695
92%   1.751
93%  1.812
94%  1.881
95%  1.960
96%  2.054
97%  2.170
98%  2.326
99%  2.576
 

Awesome, thank you sir!

Reply Post

You must log in to post a reply. Click here to login.