Speaking Notes

PADM 5502

November 5, 2009

Dr. Neubauer

WERE WE ARE

• We are in chapter 12 of the textbook regarding testing for statistical significance in the possible relationship between two variables.
• Your third (final) installment of the course term paper is due December 3.
• Please complete the online training and give me a copy of your certificate. Keep the original to be included with your IRB request for expedited review.
• IRB training link . . .

• http://phrp.nihtraining.com/users/login.php
• register with an e-mail account and so forth
• read the contents of that page and then "click here to begin"
• consider downloading the PDF version of the entire course -- on next page, right click on "Download the PDF version" and save
• to your own computer in the location of your choice.
• The first quiz is at the end of the section titled, Codes and Regulations.
• Take the quiz.  If necessary, retake the quiz.
• Take all the quizzes and print out your certificate.  Make copies of your certificate.  Give me a copy -- not the original.

Chapter 12 highlights

• Depending upon the kind of data your are using the most likely statistics to be used to measure degree of association apparent in the data is CORRELATION and Chi Square.
• In either case, there will be a "p" value associated with the statistic.
• It is the value of the p that tells you whether or not the relationship is STATISTICALLY SIGNIFICANT.
• "p" stands for probability.

If p is less than .05 you have evidence of a relationship between the two variables that is STATISTICALLY SIGNIFICANT.

You have support for your related hypothesis IF

1) you anticipated a relationship (not a null hypthesis)

2) the relationship evident in your data IS IN THE RIGHT DIRECTION.

The way to report the finding is like the following.

As shown in figure 7, men were found to be more likely than women to watch Sunday football games on television (ChiSq = 3.147, p < .05).

SO, WHY ARE SMALL p's GOOD?

The value of p is the probability that the sample obtained was randomly drawn from a population in which there is NO RELATIONSHIP BETWEEN THE TWO VARIABLES.

0.05 is five chances in 100

0.01 is one chance in 100

ASSUMING THAT YOUR SAMPLE WAS RANDOMLY DRAWN FROM THE POPULATION, the fact that you found evidence of a relationship in the sample means that there probably really is a relationship between the two variables in the population.

It is possible to draw a random sample in which there is evidence of a relationship from a population in which there is in fact no relationship. (If we put ten red ping pong balls in a hat and ten while ping pong balls in a hat, shook the hat, and then randomly drew out five ping pong balls, there is a very small change that they will be all red. A "fluke" sample is possible but not likely."

If p is less than .05 there is only a very small change that the random sample you drew from your population was a "fluke sample" and that there is in fact no relationship between the two variables in the population. THAT is the meaning of the finding being STATISTICALLY SIGNIFICANT. In other words, "I really think there is a relationship between these variables in the population because there is only a VERY SMALL CHANCE that my sample (in which a relationship is apparent) is a fluke.

---------------------

For a class exercise today . . .

1) In a Word document, write a hypothesis involving two variables -- one independent and the other dependent. Please write it in the way taught in this course. The DV should be a belief or attitude that can be measured using a Likert-style question.

2) In the same word document, write both the questions used to measure the two variables. Write them as they would appear in a SURVEY RESEARCH INSTUMENT.

3) Open Excel. In A1 cell put ID. In B1 cell put var1. In C1 put var2.

4) Enter data for 100 cases beginning in cells A2 and B2. Just make up the data for now.

5) Based on what we did in class last week, do whatever is necessary to get the data ready to be used to test your hypothesis. var2 should have a Likert response pattern. Compute newvar2, identify the missing value code, consolidate the values from 5 values to 3 values, give newvar2 a variable label and value labels (Disagree, Undecided, Agree).

6) Do whatever is necessary to prepare var1 for its use as the independent variable.

7) Use the crosstabs command to test for a relationship using the Chi Square statistic.

8) Report what you see. Is p less than .05? Do you have expected cell counts less than 5?

9) Copy and paste the two tables together into your Word document.

10) Please put your name in the Word document, save it, and send it to me as an email attachment.