Jim's Stop and Stats
Instructions & Scenario Primer
THE INFORMATION BELOW IS NOT TO BE COPIED, EDITED, OR SOLD WITHOUT THE EXPRESS, WRITTEN PERMISSION OF THE FIRST AUTHOR.
Jim L. Query, Jr., Ph.D., & Sarah Kerwin, B.A.
January, 1999 version
Title---Using a decision tree in coming to grips with statistical scenarios.
Instructions prepared for an on-line tutorial accompanying the CMUN 248 course, Loyola University Chicago, revised January, 1999. Note that a (-) that precedes content represents a hypothetical response from learners.
Opening Comments
WELCOME to the latest version of stat info and exercises for CMUN 248, Observing and Measuring Communication.
- Oh geez, life is indeed over!!
He---y, come on now. Please DO NOT admit defeat before you start the process.
- But Jim, I am no mathematician or stat guru.
That's cool...really it is...REMEMBER, you need not be a mathematical genius or Sven Gali to excel with this material. What will be involved, however, are time, diligence, repetition and application of decision rules.
- Say what?? Now you come on Jim...stats do not involve decisions.
That's alright...you persist with that notion for now...I like and encourage healthy skepticism.
Let's take off, then, by considering some fundamental strategies for figuring out research scenarios. These are addressed in a particular sequence.
- Oh please Jim...get a life dude...
Thanks for the advice...*s*...I think you'll soon discover, however, that to peruse, understand, and evaluate effectively Consumer Reports, scholarly journals/search engines, and your local paper(s), you will need to operate from an informed stat knowledge base. That's our goal here...to help bring you up to a basic level of understanding; even if you have no plans to pursue grad school.
When considering a research scenario, it is first necessary to determine which level of data, NOIR, is involved. Misidentification is like the kiss of death as everything that follows is contingent on this data type. For our purposes, the respective scenarios will only employ 1-2 types of data.
Assuming you've honed in on the correct data type(s), it is imperative to determine the nature of the inquiry. That is, are we testing for differences, or for the existence of an association?
Next, we look at the sample in terms of numbers of folks involved, number of groups involved (if any), and the values of the data. This info is essential to help us determine degrees of freedom (df). Please refer to df handout [insert location].
We still need one more piece of info: the level of probability for the hypothesis/RQ tests. Recall that we will be working with .05.
Those are the preliminary decision rules comprising the proverbial tree.
Now, wait a minute, Jim...there's got to be more than that???
There is, but only after we have analyzed the data and calculated the test statistic (i.e., the one involved in the analysis such as the t, chi-square, Pearson correlation, or Spearman rho correlation). Using the appropriate formula, we can calculate the test statistic.
And of course, most likely, a pc or canned stat package will do this for you on the job. You will still need to know what that analysis reveals though.
Then, we go to the appropriate table in the appendix: t table for t-tests; or chi table for chi squares. Armed with your calculated value (the result after using the appropriate formula), you then use the df and alpha level (.05 for example), to find the critical value.
Comparing the calculated value with the table's critical value, you are then prepared to make two decisions. Prior to those decisions, you must determine if the calculated value (from your formula and calculations) exceeds the table's critical value (these are given according to df and probability level).
We're almost done so hang tough! If your calculated value exceeds the table's critical value, you reject the null (decision one). You then conclude that in this study, your hy/RQ was confirmed (decision two).
In contrast, if your calculated value is less than or equal to the table's critical value, you embrace the null (decision one). You then conclude that in this study, your hy/RQ was not confirmed (decision two).
And yes, that is it!!! Please, don't toss any virtual tomatoes my way!!
Jim L. Query, Jr., Ph.D., & Sarah Kerwin, B.A.
January, 1999 version
Title---Using a decision tree in coming to grips with statistical scenarios.
Instructions prepared for an on-line tutorial accompanying the CMUN 248 course, Loyola University Chicago, revised January, 1999. Note that a (-) that precedes content represents a hypothetical response from learners.
Opening Comments
WELCOME to the latest version of stat info and exercises for CMUN 248, Observing and Measuring Communication.
- Oh geez, life is indeed over!!
He---y, come on now. Please DO NOT admit defeat before you start the process.
- But Jim, I am no mathematician or stat guru.
That's cool...really it is...REMEMBER, you need not be a mathematical genius or Sven Gali to excel with this material. What will be involved, however, are time, diligence, repetition and application of decision rules.
- Say what?? Now you come on Jim...stats do not involve decisions.
That's alright...you persist with that notion for now...I like and encourage healthy skepticism.
Let's take off, then, by considering some fundamental strategies for figuring out research scenarios. These are addressed in a particular sequence.
- Oh please Jim...get a life dude...
Thanks for the advice...*s*...I think you'll soon discover, however, that to peruse, understand, and evaluate effectively Consumer Reports, scholarly journals/search engines, and your local paper(s), you will need to operate from an informed stat knowledge base. That's our goal here...to help bring you up to a basic level of understanding; even if you have no plans to pursue grad school.
When considering a research scenario, it is first necessary to determine which level of data, NOIR, is involved. Misidentification is like the kiss of death as everything that follows is contingent on this data type. For our purposes, the respective scenarios will only employ 1-2 types of data.
Assuming you've honed in on the correct data type(s), it is imperative to determine the nature of the inquiry. That is, are we testing for differences, or for the existence of an association?
Next, we look at the sample in terms of numbers of folks involved, number of groups involved (if any), and the values of the data. This info is essential to help us determine degrees of freedom (df). Please refer to df handout [insert location].
We still need one more piece of info: the level of probability for the hypothesis/RQ tests. Recall that we will be working with .05.
Those are the preliminary decision rules comprising the proverbial tree.
Now, wait a minute, Jim...there's got to be more than that???
There is, but only after we have analyzed the data and calculated the test statistic (i.e., the one involved in the analysis such as the t, chi-square, Pearson correlation, or Spearman rho correlation). Using the appropriate formula, we can calculate the test statistic.
And of course, most likely, a pc or canned stat package will do this for you on the job. You will still need to know what that analysis reveals though.
Then, we go to the appropriate table in the appendix: t table for t-tests; or chi table for chi squares. Armed with your calculated value (the result after using the appropriate formula), you then use the df and alpha level (.05 for example), to find the critical value.
Comparing the calculated value with the table's critical value, you are then prepared to make two decisions. Prior to those decisions, you must determine if the calculated value (from your formula and calculations) exceeds the table's critical value (these are given according to df and probability level).
We're almost done so hang tough! If your calculated value exceeds the table's critical value, you reject the null (decision one). You then conclude that in this study, your hy/RQ was confirmed (decision two).
In contrast, if your calculated value is less than or equal to the table's critical value, you embrace the null (decision one). You then conclude that in this study, your hy/RQ was not confirmed (decision two).
And yes, that is it!!! Please, don't toss any virtual tomatoes my way!!
Jim's Research Scenarios
