Higher Ed Questons

Questions in Higher Education itself, particularly those related to mathematics and statistics education.

[higheredquestions]

My Program with 2 ANOVAs

libname mydata "/courses/u_coursera.org1/i_1006328/c_5333" access=readonly;

DATA new; set mydata.gapminder_pds; 

/* POLITY is "Democracy minus Autocracy ranges from -10 to 10" */

/*POLITY RANGES FROM -10 TO 10, making groups */

/*SETTING -10 TO -8 as 0, -7 to -5 as 1 

-4 to -2 as 2, -1 to 1 as 3, 2 to 4 as 4, 

5 to 7 as 5, and 8 to 10 as 6 */

IF POLITYSCORE=-10 OR POLITYSCORE=-9 OR POLITYSCORE=-8 THEN POL=0;

ELSE IF POLITYSCORE=-7 OR POLITYSCORE=-6 OR POLITYSCORE=-5 THEN POL=1;

ELSE IF POLITYSCORE=-4 OR POLITYSCORE=-3 OR POLITYSCORE=-2 THEN POL=2; 

ELSE IF POLITYSCORE=-1 OR POLITYSCORE=0 OR POLITYSCORE=1 THEN POL=3;

ELSE IF POLITYSCORE=2 OR POLITYSCORE=3 OR POLITYSCORE=4 THEN POL=4;

ELSE IF POLITYSCORE=5 OR POLITYSCORE=6 OR POLITYSCORE=7 THEN POL=5;

ELSE IF POLITYSCORE=8 OR POLITYSCORE=9 OR POLITYSCORE=10 THEN POL=6;

PROC SORT; by COUNTRY; 

PROC ANOVA; CLASS POLITYSCORE;

MODEL ALCCONSUMPTION=POLITYSCORE;

MEANS POLITYSCORE/DUNCAN;

PROC ANOVA; CLASS POL;

MODEL ALCCONSUMPTION=POL;

MEANS POL/DUNCAN;

run;

[higheredquestions]

MY ANOVA

Using Gapminder data, I was looking for a relationship between Polity and alcohol consumption.

Polity is defined from -10 to 10 as Democracy minus Autocracy, so -10 is least democratic and 10 is most democratic. 

The Alcohol Consumption rates are not reported for some countries, but it is the 2008 alcohol consumption rate of persons ages 15 and up (in litres pure alcohol per person) 

First I decided to group the data from 0 to 6 to have fewer categories. I called this POL (because of lack of imagination) and 0 was for -10 to -8, 1 for -7 to -5, 2 for -4 to -2, 3 for -1 to 1, etc.   

with these categories, I executed an ANOVA F-test and found with p<.001 a significant difference of the mean of the alcohol consumption of the countries with POL=0 (so least democratic) with more consumption than the others and no significant difference between the other classes. 

BUT when dis-aggregated (i.e. leaving -10 to 10 there is no significant difference between the classes). 

[higheredquestions]

ANOVA post hoc test

performing multiple tests means overall chance of type 1 error may be over 5%

many kinds of comparison tests.....

Critical Difference: LSD=Low, Tukey=Moderate; Scheffe=high

Duncan test

MEANS variable /DUNCAN;

[higheredquestions]

ANOVA output (reading)

[higheredquestions]

ANOVA

Use ANOVA F-test for C-->Q 

H0: means are equal

H1: means not all equal

F = (variation among sample means)/ (variation within groups) 

p-value = probability of this happening randomly. 

p-value is small can reject the null hypothesis and assume there is some variation between the population means. 

LIBNAME

DATA

LABEL

set appropriate missing data and create appropriate secondary variables, subset data

PROC SORT by unique identifier

PROC ANOVA; CLASS categorical_explanatory_variable;

MODEL quantitative_response_variable=categorical_explanatory_variable;

MEANS categorical_explanatory_variable;

run;

[higheredquestions]

Notes on graphing 2

Bivariate data relationships: explanatory (independent) variable and Response (dependent) variables. 

manipulation of categorical variables. 

[higheredquestions]

Notes on Graphing 1

Bar charts: frequency and relative frequency of categorical variables

SAS: 

Note you can "comment out" (i.e. hide the syntax you want to ignore).

Use Tasks -> graphs -> bar chart wizard. 

MISTAKE IN THIS LESSON! Discrete can be quantitative as in when you count something. But continuous is also quantitative. 

Hint may need to run the program once before graphing. 

Graph bar charts and histograms. then talk about center, shape and spread, measures of center and dispersion. 

SAS: 

after PROC SORT:

PROC UNIVARIATE;  VAR quantvariable1; 

[higheredquestions]

Notes on Data Management

Step 1: Do you need to code out missing data?

IF variable=code_for_missing THEN variable=.;  /* Dot for missing data */

(Note: all data management statements go below the DATA statement and above the PROC SORT statement. And before any data logic statements that subset data to specific observations.)

Step 2: Do you need to add valid data? 

Look for skip patterns for the NA.

IF variable=. THEN variable=dummy_code;

Step 3: Give variables values that are more logical. 

Reverse the codes so small means small using new variables and document well in the program. Or recode so that categorical becomes quantitative, but retains order. 

(STOP AT 12:33)

[higheredquestions]

SAS notes

From video 1:

You can label the data immediately after the DATA statement using:

LABEL variablename1="Label I want to use for variable1" variablename2="Label I want to use for variable 2";

You can choose particular data values to use (such as age <= 25 and smoking only in last 12 months) by using   logic statements just before PROC SORT as listed below.

The instructor wanted to have age  used:

IF CHECK 321=1;

IF AGE LE 25; 

She also added age to her table statement. 

Comments are added between /* and */  

[higheredquestions]

Not using those variables?

Well, it looks like we can only use the variables in the codebooks. 

for now, I'l do that. But I don't have any PASSION for those, so hope we can learn how to use others....

so I'll play with gapminder looking at polityscore and suicideper100TH 

my program:

DATA new; set mydata.gapminder_pds; 

PROC SORT; by COUNTRY; 

PROC FREQ; TABLES polityscore; 

run;

and my output:

[higheredquestions]

Notes from "Working with Data"

EDA= Exploratory Data Analysis, organizing and summarizing data, looking for patterns (any striking deviations) and interpreting the findings in context of the data. 

Time to go: Start SAS Enterprise. (2:43 -  Credentials required and login not working....)

ok.all set now. 

to write a program, select: file-->new-->program

Now the statements:

LIBNAME: call in a data set. 

from the discussion board: 

libname mydata "/courses/u_coursera.org1/i_1006328/c_5333" access=readonly;

(name, sascloudname, don't modify raw data) MUST end with a semicolon

DATA: reads data set and prepares it for use.

 DATA new; set mydata.nesarc_pds;

SET: tells SAS which data file to set for analysis. (see the libname above, file to use. for gapminder use gapminder_pds)

The instructor suggests to next use:

PROC SORT: to end the data step. asking SAS to sort the data by the unique identifier from the variable in the data set fro gapminder it is COUNTRY 

PROC SORT; by IDNUM; 

Now save the program using: SAVE or SAVE AS in the menu. 

To Create a Frequency Distribution use:

PROC FREQ; TABLES names of variables separated by blanks;

run; 

then save and run. 

[higheredquestions]

Assignment one

codebook = data dictionary

Choose a data set and obtain the codebook.

I will be using Gapminder. 

Select topic and related variables.

My topic is educational attainment and I'm interested in Mathematics and girls.  

Variables: 

        Select a research question.

How does the mathematical attainment correlate with the Human Development Index?

Identify a second topic associated with original question and select variables. 

How does the ratio of boys and girls in school correlate with the HDI? 

Variables given above. 

[higheredquestions]

Lecture 1

Use "categorical" over Qualitative? Is this common across the disciplines? 

[higheredquestions]

The essence of mathematics is not to make simple things complicated, but to make complicated things simple.

S. Gudder

[higheredquestions]

First Post

This is my first post for the Coursera March 2013 Course: Passion Driven Statistics.  Bonus points if you recognize the avatar. Hint: I took a photo of it in Boston, MA, USA.  

I am Susan, a Math Professor at a Community College in Massachusetts. I've been teaching Introductory statistics for about 3 years. And I have never taken a statistics course myself.