SPSSLab

Gail=s

SPSS Lab Workshop

Winter 2002

Research Methods

Gail Johnson

The Evergreen State College

Masters of Public Administration Program

867-6739

johnsong@evergreen.edu

Table of Contents

Purpose and Objectives: 1

Basic Concepts: 1

SPSS Overview: 2

The Data Sets Used in Course 3

Tasks 4

Opening a Data Set:

Creating a Data Set

Entering Data

Saving and Close

Analyzing Data 11

Frequencies: 11

Central Tendency: Mode, Median, Mean. 13

Measures of Dispersion and Standard Deviation 14

Cross Tabs(contingency tables) 14

Comparing Means 16

Confidence Intervals: 18

Statistical Significance: 19

1. Chi-Square: Tests for Nominal and Ordinal Data: 19

2. T-Test: Testing for Statistical Significance for Interval and Ratio Data: 21

3. One-way Anova (Analysis of Variance).......................................................... 23

Relationship: Measures of Association 24

Simple Regression: 27

Multiple Regression: 28

Mananaging Output.............................................................................................................................30

Moving SPSS Tables to Documents 31

Graphics 32

SPSS Workshop Guide

Purpose and Objectives:

The purpose of workshop is to provide hands-on experience in using SPSS.

Note: you may find it helpful to try the tutorial on your own at a later time.

These exercises are designed to familiarize you with the following:

# Opening a Data Set

# Adding variables to a data set

# Saving a Data Set

# Analyzing a Data Set .

Working with Computers: General Advice

( Approach the computer with a calm mind.

You can=t break the machine, unless you drop it on the floor!
Mistakes are part of the learning process, so do not be afraid to try things
You will figure it out---it just takes some time and patience
Don=t work under pressure!
If you get frustrated, take a break.
Try things three times Bif you can=t get it, ask for help.
Play with the software and the data. Explore! No one is watching to see what you are doing.

Basic Concepts:

Data Set: A specific set of data that is in a file format which resembles a spread sheet with rows and columns. The rows contain the information for each person or case; this is called a Arecord.@ The columns contain the specific information for each person: age, gender, income, etc. Every record will have a unique identifier, usually called an ID.

Variables: These are the specific items which are collected and entered into the columns of a data set. For example in the typical survey data set, age, gender, race, income, etc. would be considered variables. Each variable has a unique name (less than 9 characters). E.g. V8

Variable Labels: Variable labels are closer to the English language, so when it prints out, it makes sense to you e.g. V8 is labeled PoliticalView.

Values: Each variable is comprised of values. If the variable is gender, then the values are male and female. In a data base, however, values are typically coded to take up less room. So gender might be coded as M or F, or gender could be coded as 0 or 1.

Value Labels: These labels show up on your output. So, in our gender example, instead of having 0s and 1s on the output, the computer will print out the value labels of Male and Female. You can create and change value labels whenever you want.

SPSS Overview:

SPSS looks like a spread sheet with rows and columns. Every row represents an individual record. Like other windows application, SPSS is menu driven. Typically, the menu is at the top of the screen.

Menu Options:

File: To open, save, or close a data set.

Edit: To cut, copy, paste, find data or text. Options: you can make changes to

how things appear. For exmaple, you can change how the variables appear on the lists in the analysis boxes: names or labels, alphabetically or in order of the file.

View: Allows you to change how you view the screen. i.e. make it larger or smaller.

Data: To insert cases or variables.

Transform: Main one we will use. To recode variables; this is useful when you want to change interval level data, such as age, into categories (18-25, 26-35 etc.).

Analyze: This is were all the fun takes place. This menu contains all the tools you will need to crunch numbers and analyze data.

Graphs: This menu will allow you to create a variety of charts, line, bar, and pie, from your data.

Utilities: This allows you to view variable labels, variable definitions, and data set Information.

Help: Help.

You will also find that right clicking the mouse will give you options. You can cut and paste to make creating data sets less labor intensive.

Determining How Variables are Coded:

When you are using a data set you did not create, you can check to see how variables are coded by double clicking on the column header. This will take you to variable view. Or you can click on the variable view at the bottom left side of screen.
In the analysis boxes, highlight the variable, right click the mouse, and then click on variable information. There will be an arrow key on the right; click on that to show the values and value labels. Note: if the data is really interval, there will be no value labels.
You can also use the utility option on the menu.

Output: Whenever you analyze some data, SPSS will create an output file. Each set can be copied, cut, and pasted into other programs such as Word or Word Perfect much like you would text or clip art.

You can also save the output to a disk or to the hard drive.

You can print out the output, as you would any document.

Warning: Each time you analyze data, the new analysis is added to the output file. When you tell the computer to print, all the output will print out. What this means is that if you have stuff you do not want, delete it before printing to save trees.

Data Sets

You have been given a disk with data sets on it: Put your name and phone number on the disk. The following data sets are to be used for your assignments:

USPA Fall 2000: All the students in USPA classes were asked to complete a student survey in Fall 2000. A copy of the survey is on your Lab disk. Some variables have been deleted so data set has <50 variables. It is similar to the 1997 survey in Appendix I in the Practical Guide. You will use this dataset for your data analysis paper.

File: USPA2000.sav

GSS Surveys : These are short version of the General Social Survey which is a random sample of U.S. Adults. These data sets include some demographic information as well as respondents attitudes on a variety of political issues, and music preferences. These set contains mostly categorical and ordinal data, but there are few variables with interval level data.

Files: gss91pol.sav and gss93.sav

Welfare Data: This data set contains mostly interval level data. Information from all 50

states include: poverty data, AFDC benefits for a family of three, teen pregnancy, etc.

File: welfare.sav

The Workshop Begins:

1. Opening SPSS

Steps:

Go into the start menu
Click on SPSS application

2. Opening an Existing Data Set

Steps:

Click File --> Open --> data.
Select Drive - a:
Select >workshop.sav= file.
Click >OK=.

3. Creating a Data Set

Now you can practice creating a data set on the Aworkshop.sav@ file. Someone started to set up this data set but did not finish. These are the survey questions on page 6. You will finish creating this data set. Your first task is to create the remaining variables (with variable names, variable labels and value labels).

Steps:

Open the workshop.sav file

Click on View Data button on in the bottom left or double click on the top of the column. This will take you to the variable data view. The variables will be listed on the left going down, and the definitions will bein the rows.
Go down the left column until you get to the row below teach9. We will be beginning here to finish setting up the data set.
Type in variable name-- teach10 (no spacesBSPSS will have heartburn).
Move to the right into the Type column. Change the decimal place from 2 to 0 for this exercise.
Move to the labels column. Type in Research Papers (this is the variable label)
Move to the Value column. It will say None. Click on the little box to the right. It will take you to a pop-up box.Type 1 in the value box under the value labels section.
Type >Very Effective= in the value label box , then click on Add button (this is the value of the variable).
Continue defining the rest of the value labels for that variable (2= Generally effective, 3=Neither etc.,) Click on Add after each entry
When you have finished labeling all the values, click OK.
Go the missing values column. It will say none. Click on the little box, it will pop-up another box. Click on discrete missing values, type 9 in the first box (if some survey responses are missing the answer, you can use 9 as a missing value), type 6 (value of >not applicable= - answer)in the next box; (defining missing values and >not applicable= answers will help you to summarize valid response while analyzing data) and click OK
Go to the Measures column. This allows you to identify whether the data is Nominal, Ordinal, or Scale (meaning interval or ratio). This is ordinal data. The world won=t fall apart if you don=t label this.
OK, this variable is set to go.

Now you completed creating a variable label and value label for one variable, move onto next questions, and finish creating the variables. Instead of just using var1, var2 etc., try to give them names that will help you during data analysis time. Age, race, program, etc. might be useful. If you use an English name as your variable name, you don=t need to label it.

Note: number of credit hours is a real number. You do not label the values for real numbers: they stand as they are.

Also note: You might want to add missing values to Teach1BTeach9 variables. You can either type in 9s and 6s as the discrete values for each one or you can:

go to missing values for Teach10 which you have already done.
Right click and copy.
Then highlight missing values for Teach1BTeach9, right click and click paste. Voila! They are all done.
File==>save

4. Entering Data

Once you have set up your data set, you are ready to enter the data. You will find 4 surveys on pages 7-10 Please enter the data into the data set.

Go back to the bottom left on your screen, anc click on data view. You can use the Atab@ key or the arrow key to move around the cells in the data base. Hit the Aenter@ key or tab to the next column once you have typed in the appropriate value.

It is often easiest to work in pairs for this, with one person calling out the data and the other person typing it. It is easy to make errors in typingBtrust me, I know. Go back and verify all your entries.

5. Saving and Closing a Data File

Steps:

Click File --> Save

Or you want to save it as a new file use Save as.

Type in a new a file name in the director you want(you can save this file as 'workshop2.sav' file).
Select Drive a: and then click 'OK'.
If want to go to another SPSS data set, Go to File B> Open and click on the file you want. SPSS will ask if you want to save your current one. Say yes if you want to save it.
If you want to get out of SPSS, you can hit the X button, top right.

Question: 3. Effectiveness of Instructional Approaches:

Considering the USPA courses, you have taken to date, how effective or ineffective are each of the following instructional approaches in helping you learn the material?
Instructional Approaches	Very Effective 1	Generally Effective 2	As Effective as Ineffective or Neither 3	Generally Ineffective 4	Very Ineffective 5
a. Lectures
b. Guest speakers
c. Case studies
d. Videos
e. Class discussions
f. Individual student presentations
g. Group student presentations
h. Group student projects
i. Exams
j. Research papers

Question: 11. What is your present age group?

_____21 to 25 _____26 to 35 _____36 to 45 _____ 46 or over.

Question: 12. What is your gender?

_____ Male _____ Female.

Question: 16. In which program are you currently enrolled?

____MPA _____MUS ____Ph.D., ____Other Program(MS/MA) ____Not enrolled.

Question: 17. Including this semester, approximately how many credit hours have you completed?

________ credit hours.

ID: 01

Question: 3. Effectiveness of Instructional Approaches:

Considering the USPA courses, you have taken to date, how effective or ineffective are each of the following instructional approaches in helping you learn the material?
Instructional Approaches	Very Effective	Generally Effective	As Effective as Ineffective or Neither	Generally Ineffective	Very Ineffective
a. Lectures				x
b. Guest speakers		x
c. Case studies	x
d. Videos		x
e. Class discussions	x
f. Individual student presentations			x
g. Group student presentations			x
h. Group student projects	x
i. Exams				x
j. Research papers	x

Question: 11. What is your present age group?

_____21 to 25 _____26 to 35 _____36 to 45 __x___ 46 or over.

Question: 12. What is your gender?

_____ Male ___x__ Female.

Question: 16. In which program are you currently enrolled?

____MPA _____MUS __x__Ph.D., ____Other Program(MS/MA) ____Not enrolled.

Question: 17. Including this semester, approximately how many credit hours have you completed?

_____36_____ credit hours.

ID: 02

Question: 3. Effectiveness of Instructional Approaches:

Considering the USPA courses, you have taken to date, how effective or ineffective are each of the following instructional approaches in helping you learn the material?
Instructional Approaches	Very Effective	Generally Effective	As Effective as Ineffective or Neither	Generally Ineffective	Very Ineffective
a. Lectures	x
b. Guest speakers		x
c. Case studies			x
d. Videos		x
e. Class discussions	x
f. Individual student presentations				x
g. Group student presentations					x
h. Group student projects			x
i. Exams		x
j. Research papers		x

Question: 11. What is your present age group?

__x___21 to 25 _____26 to 35 _____36 to 45 _____ 46 or over.

Question: 12. What is your gender?

___x__ Male _____ Female.

Question: 16. In which program are you currently enrolled?

__x__MPA _____MUS ____Ph.D., ____Other Program(MS/MA) ____Not enrolled.

Question: 17. Including this semester, approximately how many credit hours have you completed?

___9_______ credit hours.

ID: 03

Question: 3. Effectiveness of Instructional Approaches:

Considering the USPA courses, you have taken to date, how effective or ineffective are each of the following instructional approaches in helping you learn the material?
Instructional Approaches	Very Effective	Generally Effective	As Effective as Ineffective or Neither	Generally Ineffective	Very Ineffective
a. Lectures	x
b. Guest speakers
c. Case studies	x
d. Videos			x
e. Class discussions		x
f. Individual student presentations					x
g. Group student presentations				x
h. Group student projects			x
i. Exams				x
j. Research papers		x

Question: 11. What is your present age group?

_____21 to 25 _____26 to 35 __x___36 to 45 _____ 46 or over.

Question: 12. What is your gender?

_____ Male ___x__ Female.

Question: 16. In which program are you currently enrolled?

___x_MPA _____MUS ____Ph.D., ____Other Program(MS/MA) ____Not enrolled.

Question: 17. Including this semester, approximately how many credit hours have you completed?

_____9_____ credit hours.

ID: 04

Question: 3. Effectiveness of Instructional Approaches:

Considering the USPA courses you have taken to date, how effective or ineffective are each of the following instructional approaches in helping you learn the material?
Instructional Approaches	Very Effective	Generally Effective	As Effective as Ineffective or Neither	Generally Ineffective	Very Ineffective
a. Lectures	x
b. Guest speakers		x
c. Case studies
d. Videos				x
e. Class discussions					x
f. Individual student presentations					x
g. Group student presentations				x
h. Group student projects		x
i. Exams	x
j. Research papers	x

Question: 11. What is your present age group?

____x_21 to 25 _____26 to 35 _____36 to 45 _____ 46 or over.

Question: 12. What is your gender?

____x_ Male _____ Female.

Question: 16. In which program are you currently enrolled?

____MPA ___x__MUS ____Ph.D., ____Other Program(MS/MA) ____Not enrolled.

Question: 17. Including this semester, approximately how many credit hours have you completed?

_____120___ credit hours.

Analyzing Data: Overview

Nominal: This is data which has a name, i.e. a person=s job, their sex, the city they live. Nominal data should not be averaged.

Ordinal: This is data which has an order (low to high). Survey data is frequently ordinal data: how strongly do you agree or disagree . Gail=s advice: ordinal data should not be averaged.

Interval and Ratio: SPSS calls these Scale. These measures are real numbers, such as: age, weight, dollars. Interval and ratio data can be analyzed using the same statistical procedures.

Descriptive Analysis

A. Using SPSS to get Frequencies and Percent Distributions:

SPSS Exercise 1: What percent of the respondents from the Fall 1997 survey were male or female?

Steps:

File == > Open ==> Data

Select students.sav

Click on Analyze from top menu
Click Describe Statistics

Click on Frequencies

On this new screen you will see a list of variables running down the left side of the screen and empty box on the right.
Click on the variable or variables your want to analyze (you can highlight several at a time) and then hit the arrow key $ . Select sex

Note: If you want to delete a variable that you have selected for analysis, click on that variable, and then hit the arrow key ³ which is pointing in the direction of the variable list on the left side of your screen. The arrow changes direction based on which box you are working in.
Click OK

SPSS Exercise 2: What is the frequency and percent distribution for the number of credits reported in the Fall 1997 survey?

Steps:

Click Analyze from menu bar
Click Descriptive Statistics

Click Frequencies.
In >Frequencies= dialog box select >credit= and click the small arrow button in the middle and move it to >variables= box.
Click OK

Output # 2:

CREDITS Credit hours completed

	Values	Frequency	Percent	Valid Percent	Cumulative Percent
Valid	0	32	20.5	20.5	20.5
	3	16	10.3	10.3	30.8
	6	16	10.3	10.3	41.0
	7	1	.6	.6	41.7
	9	15	9.6	9.6	51.3
	12	12	7.7	7.7	59.0
	13	1	.6	.6	59.6
	15	11	7.1	7.1	66.7
	18	9	5.8	5.8	72.4
	21	6	3.8	3.8	76.3
	22	2	1.3	1.3	77.6
	24	1	.6	.6	78.2
	25	1	.6	.6	78.8
	26	2	1.3	1.3	80.1
	27	3	1.9	1.9	82.1
	30	7	4.5	4.5	86.5
	33	4	2.6	2.6	89.1
	34	1	.6	.6	89.7
	36	5	3.2	3.2	92.9
	39	2	1.3	1.3	94.2
	40	1	.6	.6	94.9
	43	1	.6	.6	95.5
	45	2	1.3	1.3	96.8
	49	2	1.3	1.3	98.1
	60	1	.6	.6	98.7
	153	1	.6	.6	99.4
	190	1	.6	.6	100.0
	Total	156	100.0	100.0

Note:

Output # 2. It will be long because it will tell you how many people (and what percent) for each value of credit hours.

How to Interpret the Output:

The second column contains the values for the variable that were reported. In this table, it is the exact number that they reported. You might note that there are no missing values in the table. The next column is the frequency column, which shows how frequently the various credit hour was selected. For example: 32 people reported they had not completed any credit hours and 12 reported they had completed 12 credit hours. The next is the percent column. This shows the percent distribution of all the people who were included in the survey. 21% reported 0 credit hours and 8% reported 12 credit hours.

The next column shows the valid percent, that is, of the people who gave an answer. In the previous example, there were missing values. In this example, everyone answered the question, so the percent and valid percent columns are the same. The last column, the cumulative percent column, adds the percent as you got down the column. 21% reported 0 credits. 10% reported 3 credit hours, which gets added to the 21%, for a total of 31%. If we want to see where the majority are, we find 51% reported having 9 credits hours or less. Majority is always 51% or more.

Extra Exercise:

1. Are students more likely to prefer lecture classes (teach1) as compared to class discussions (teach5)?

2. Are students more likely to prefer exams (teach 9) as compared to research papers (teach 10)?

3. What percent of students would recommend the program (recommnd)?

B: Using SPSS to Get Means, Medians and Modes

Exercise 3: How would you best describe central tendency of the number of credit hours of the respondents?

Steps:

Click on Analyze ==> Descriptive statistics ===> Frequencies

Select the variable (or variables) you want from the box on left side of the screen and click to highlight. Credits

Click on the arrow key to move them to the right box.
At the bottom of the screen, click on statistics.
You will be taken to a new screen where you can select mean, median, and/or mode by clicking whichever function you want.
Click Continue to go back to original box.
Hint: If you are working with interval level data and don=t want a long table of output to run, while you are still in the frequencies box, click on the bottom button that says format. You then click on the button that says suppress tables if more than 10 values.
Click OK again.

Output 3: Credit hours:Statistics

CREDITS Credit hours completed

N	Valid	156
	Missing	0
Mean		15.58
Median		9.00
Mode		0

C. Using SPSS to Get the Standard Deviation

You have options for doing this. You can use the options in the frequency window, as you did above.

Exercise 4: What is the dispersion of credit hours?

Steps:

Click on analyze

Click on Descriptive Statistics,
Click on frequencies

Select your variable (credit) and move to right box.
At the bottom of the window, click on statistics

Click on standard deviation, range, and minimum and maximum.

Hit OK

We can use a different analysis on SPSS to get the standard deviation: Descriptive command. The choice is a preference thing.

Steps:

Click Analyze --> Descriptive Statistics --> Descriptive.
Select the variable >credits= and move it to >variable box on right
Click on Options button at bottom of window
Select Std. Deviation, Minimum, Maximum
Click Continue.

Click >OK=.

D: The Concept of Crosstabs (contingency tables): Used with Nominal and Ordinal data.

Cross tabs are perhaps the most interesting and fun statistical tool you can use. It can be used to describe two or more variables, as well as examine relationships.

Dependant Variable: This is the variable you want to explain.

Independent Variable: This is the variable you think affects your dependant variable or causes it to change.

Using SPSS to Do Crosstabs:

Exercise 6: Suppose we want to see if gender matters in terms of extent to which respondents will recommend the program to others.

Steps:

Click on Analyze on the tool bar
Click on Descriptive Statistics
Click on cross tabs.
A new window will pop up. As before, you will highlight the variables you want to analyze and click the appropriate arrow.
Select recomnd (Extent of recommend) and move ==> to the column box. This is our dependent variable.
Select sex and move==> to the row box. This is our independent variable.
Click on cells at the bottom of the window.
In the new screen that pops up, go to percentages and click on percentage on row. This is how we want to analyze this data.
Click on continue to take you back first window.
Click OK

Note: you always percentage the row or column where you place your independent variable.

Output # 6:

SEX Sex * Extent of recommendation Crosstabulation

	Extent of Recommendation						Total
			Strongly	Generally	Possibly	Generally not
SEX Sex	Male	Count	17	26	7	1	51
		% within SEX Sex	33.3%	51.0%	13.7%	2.0%	100.0%
	Female	Count	19	20	7	1	47
		% within SEX Sex	40.4%	42.6%	14.9%	2.1%	100.0%
Total		Count	36	46	14	2	98
		% within SEX Sex	36.7%	46.9%	14.3%	2.0%	100.0%

Interpreting the Output:

Of the 51 men, 33% would strongly recommend and 51% would generally recommend. Now look at the female row. How many women are there? 47. What percent of them strongly recommended the program? 40% strongly recommended and 43% generally recommended the program.

Gail=s advice: I look for at least a 10% difference between the results. In this case, I would look for at least a 10% difference before I am willing to say there is a difference in extent to which men and women would recommend this program. These results do not have a 10% difference.

Exercise 6A: Let=s Review this setting up the tables stuff:

You can set up the table differently. You can put sex in column and recomnd in the row. When you go to cells, you want to click on percentage in column, since that is where you placed the independent variable.

Steps:

Click on Analyze on the tool bar
Click on Descriptive Statistics
Click on cross tabs. You will note it has your last analysis.
Select recomnd (Extent of recommend) and hit arrow key to put it back on the variable list box. Then hit the arrow to move it to the row box. This is still our dependent variable.
Select sex and move it back to the variable list, and then click the arow to move it to the column box. This is still our independent variable.
Crucial to remember: Click on cells at the bottom of the window.
In the new screen that pops up, go to percentages and click on percentage on column. This is how we want to analyze this data.
Click on continue to take you back first window.
Click OK

Even more confusing: The computer will do whatever you ask. So what does it look like if we click on other cell options?

Exercise 7: Go back to your analysis on sex and recommendation. Go through the steps to set up the cross tab analysis, but when you go into cells, click on the options for percent: expected, row, column, total. Print this out and bring this output to class. We will go over this mondo bizarro table!!

Exercise 8:

a. Do people from different programs (program) have more or less similar in their willingness to recommend (recommnd) the program?

B. Do students vary in their preference for research papers (Teach10) and exams (teach9) based on what program (MPA, MUS or Ph.D) they are in?

E. Using SPSS to Compare Means:

In situations where your dependent variable is interval/ratio level data, and your independent variable is nominal, you can compare the means.

Exercise 9: Do men earn more than women?

To answer this question, we need to use a different data set.

You need to open a new data set: GSS91pol.sav

Steps:

Click on File==> Open===> A===> GSS91pol.sav

Click on Analyze

Click on Compare Means

Click on Means

This will pop-up the box. One the left will be the list of variables, and on the right, will be 2 boxes: Dependent and Independent.
Select on respondents sex and move to the independent box.
Select on incomdol (income recoded to dollars) and move to the dependent box.
Click on Options button
Select Mean, Std. Deviation, Median.
Click Continue

Click OK

Output 9

Income recoded to dollars

Respondent's Sex	Mean	N	Std. Deviation	Median
Male	38965.11	609	25470.86	32500.00
Female	33096.23	756	25365.95	27500.00
Total	35714.65	1365	25570.58	32500.00

Exercises 10: Compare means using the GSS91pol data:

a. find out if men and women watch the same amount of TV (TVhrs).

b. find out if there is a difference in the age of marriage (agewed) between men and women.

Exercise 11: Is there is a difference in income (income recoded as dollars: name incomdol) and educational degree (RS Highest Degree?

Output # 11:

Report

Income recoded to dollars

RS Highest Degree	Mean	N	Std. Deviation
Less than HS	18021.08	249	18217.58
High school	33188.21	704	22530.35
Junior college	41129.31	87	22645.84
Bachelor	49033.56	216	25542.65
Graduate	62275.46	108	25168.22
Total	35738.27	1364	25565.06

Concept of Inferential Statistics: Working with Sample Data

A. Confidence Levels and Confidence Intervals.

Sometimes we want to make estimates about the larger population based on our sample. Because we are working with a sample, there is always some amount of error. But we still want to be able to say with some degree of confidence (usually 95% confident) that the average income is at least between two points. One way to estimate averages back to the larger population is by constructing a confidence interval. Typically, you will ask the computer to tell you, with 95% confidence, what the true mean would be if we had surveyed everyone in the population.

Using SPSS to Calculate Confidence Intervals:

Exercise 12: What is the best estimate of the average salary of the population, and the 95% confidence interval, based on our sample?

Steps:

Click on Analyze ==> Describe Statistics ===> Explore

Select the variable Incomdol ( income recoded to dollars) and move it into the Adependent box@

Click OK

Output 12:

Descriptives

		Statistic	Std. Error
Mean		35714.65	692.11
95% Confidence Interval for Mean	Lower Bound	34356.94
	Upper Bound	37072.36
5% Trimmed Mean		34682.13
Median		32500.00
Variance		653854367.273
Std. Deviation		25570.58
Minimum		500
Maximum		87500
Range		87000
Interquartile Range		41250.00
Skewness		.714	.066
Kurtosis		-.516	.132

Exercise 13: Based on our sample:

What is our best estimate of average number of siblings in the population?

What is our best estimate of the average age that people in the population wed?

Print out and Bring to Class.

Another Analysis: You can use explore to obtain CI if you are comparing means.

Exercise 14: Suppose you want to know the means and the confidence intervals of the average salaries for men and women?

Steps:

Click on Analyze ==> Describe Statistics ===> Explore
move incomedol to dependent box

move sex to factor box

Click OK

Output 14:

Descriptives: Income for Men and Women

Respondent's Sex			Statistic	Std. Error
Male	Mean		38965.11	1032.13
	95% Confidence Interval for Mean	Lower Bound	36938.13
		Upper Bound	40992.08
	5% Trimmed Mean		38254.70
	Median		32500.00
	Variance		648764713.845
	Std. Deviation		25470.86
Female	Mean		33096.23	922.55
	95% Confidence Interval for Mean	Lower Bound	31285.16
		Upper Bound	34907.30
	5% Trimmed Mean		31792.99
	Median		27500.00
	Variance		643431538.750
	Std. Deviation		25365.95

B. Concept of Statistical Significance:

When you are working from a random sample you must figure out if somehow you got your results by chance. If the results are .05 or less, we will say the results are statistically significant and therefore results we observe are not likely to be caused by random chance.

1. Using SPSS to Do Chi-Square Tests for Nominal and Ordinal Data:

To do a chi square, you first do the cross tab analysis.

Exercise 15: Is there a statistically significant difference between men and women willingness to recommend the program.

Steps

Analyze ==> Descriptive Statistics ===> Crosstabs

Select gender and move to row

Select recommend program (recomnd) and move to column.
Click on cell and select percent on row.
Click Continue

New step: Click on Statistics

Click on Chi Square

Click on Continue

Click OK

Note: this is Chi Square for the crosstab you did earlier that answered whether or not men and women differed in there responses on recommending the program (Exercise 6, output 6). As you recall, we did not see much of a difference. The Chi Square Table looks like this:

Output 15:

Chi-Square Tests

	Value	df	Asymp. Sig. (2-sided)
Pearson Chi-Square	.732	3	.866
Likelihood Ratio	.733	3	.865
Linear-by-Linear Association	.136	1	.713
N of Valid Cases	98

Interpreting the Output:

Remember, we are pretending this is a random sample for a large population of students for this exercise.

The only column of importance is the last one: Sig. (2 sided). For the results to be statistically significant, the test should be .05 or less. Clearly, .866 is greater than .05. The relationship between sex and extent to which respondents recommended the program is not statistically significant.

Exercise 16: Is political outlook (Liberal/conservative) (politics) related to views on legalization of marijuana (grass)?

Open the GSS91 Pol data set. Do a cross tab, with Chi Square test for statistical significance. Remember, to click cell and percentage on wherever you place your independent variable.

2. T-Test: Testing for Statistical Significance for Interval and Ratio Data:

The T-Test is an interval statistic that determines if the means are statistically significant.

One Sample: Use when you are comparing sample results to a known population.

Paired Means: Use when two variables are paired like in a pre and post test: before and after design. You must use the same people for your results to be correct.

Independent Means: Use when you are comparing the means of two groups (for example, men and women)..

Using SPSS to Do a One Sample T-Test

Exercise 17: If you were trying to figure the number of hours worked per week and whether the results you got were pretty accurate of the population in general. Use GSS91pol.sav

One Sample T-Test: Steps:

Analyze ==> Compare Means

Click One Sample T-test

Select hours worked (HRS1) as your variable
Select a test level. In this case we would expect 40 hours to be the norm.
Click OK

Output 17:

One-Sample Statistics

	N	Mean	Std. Deviation	Std. Error Mean
HRS1 Number of Hours Worked Last Week	900	41.76	14.63	.49

One-Sample Test

	Test Value = 0
	t	df	Sig. (2-tailed)	Mean Difference	95% Confidence Interval of the Difference
					Lower	Upper
HRS1 Number of Hours Worked Last Week	85.636	899	.000	41.76	40.81	42.72

Using SPSS to Do an Independent Sample T-Test:

Let=s go back to our gender question about income. In prior analysis, we found that men earned more, on average, than women (Exercise 9).

Exercise 18: Are these differences in income statistically significant? We use an independent sample t-test.

Steps:

Analyze ==> Compare Means

Click on Independent T-test

Select Income (incomdol) and move to dependent box
Select Sex and move to independent box.
Click >Define Groups= Button, type >1' in >Group 1' box and type >2' in >Group 2' box. You need to tell the computer that 1 means male and 2 means female, if that has the data was coded. To check, you can right click the mouse after you highlight the variable and click on variable information.

Click OK

Output 18:

Group Statistics

Respondent's Sex	N	Mean	Std. Deviation	Std. Error Mean
Male	609	38965.11	25470.86	1032.13
Female	756	33096.23	25365.95	922.55

Independent Samples Test

	Levene's Test for Equality of Variances		t-test for Equality of Means
	F	Sig.	t	df	Sig. (2-tailed)

Equal variances assumed	.474	.491	4.241	1363	.000
Equal variances not assumed			4.239	1299.580	.000

Interpreting the Output:

We get two tables. The first table tells us the mean and standard deviation. The second gives us the statistical test for significance. This is presented in the last column, Sig. 2-tailed. The statistically test is less than .05. We can ignore all the rest of this stuff.

Exercise 19:

How likely are the average number of hours spent watching TV (TVhours) reported by our respondents a fairly accurate portrayal of the population as a whole?

Are there differences in the average number of hours of TV watched by men and women, and are those results statistically significant?

Analysis of Variance:

You may have noticed that the t-test grouping variable only accepts two values. When you have a variable with more than two values, such as race, religion and educational degrees, you will need to use one-way Anova (analysis of variance) to test for statistical significance.

Exercise 20: Is there a statistically significant difference in income based on educational degree (highest degree obtained)?

Steps:

Analyze ==> Compare Means

Click on one-way anova

Select Income (incomdol) and move to dependent box
Select RS educational degree (degree) and move to factor box.
Click on options and select descriptive.

Click OK
Click OK

Output 20:

Descriptives

Income recoded to dollars

	N	Mean	Std. Deviation	95% Confidence Interval for Mean		Minimum	Maximum
				Lower Bound	Upper Bound
Less than HS	249	18021.08	18217.58	15747.22	20294.94	500	87500
High school	704	33188.21	22530.35	31521.05	34855.37	500	87500
Junior college	87	41129.31	22645.84	36302.83	45955.79	2000	87500
Bachelor	216	49033.56	25542.65	45607.95	52459.18	500	87500
Graduate	108	62275.46	25168.22	57474.50	67076.42	5500	87500
Total	1364	35738.27	25565.06	34380.35	37096.19	500	87500

ANOVA

Income recoded to dollars

	Sum of Squares	df	Mean Square	F	Sig.
Between Groups	199504503974.035	4	49876125993.509	98.048	.000
Within Groups	691314308342.681	1359	508693383.622
Total	890818812316.716	1363

VII. Relationships

Sometimes it is important to determine how strong the relationship is between variables. Measures of relationships are generally measured from zero to one, with 1 being a perfect relationship and 0 being no relationship at all. Most of the time you will see the relationship reported in a number which has a plus(+) or a minus(-) sign next to it. These signs refer to whether the relationship is an inverse(-) relationship, where as one goes up the other goes down, or a positive(+) relationship, where the variables change in the same direction. As one variable goes up so does the other or as one goes down so does the other.

Measures of Association:

Using SPSS to Measure Relationships for Nominal Data:

Exercise 21: Is one=s religious preference related to one=s veiws on programs to provide birth control to teenagers? Use the GSS91pol data set.

Steps:

Analyze ==> Descriptive Statistics ====> Crosstabs

Select Relig (Religious Preference) and move to row. (This will be our independent variable)
Select PillOK (Birth Control to teens) and move column
Click on cells to do your percentage on row

Then click on continue.

Click on the statistics at the bottom of this box.
Go to Nominal, and select on Phi / Cramers V.
Select Chi Square too.
Click OK

Chi-Square Tests

	Value	df	Asymp. Sig. (2-sided)
Pearson Chi-Square	32.672	12	.001
Likelihood Ratio	35.095	12	.000
Linear-by-Linear Association	22.588	1	.000
N of Valid Cases	970

a 6 cells (30.0%) have expected count less than 5. The minimum expected count is 3.17.

Symmetric Measures

		Value	Approx. Sig.
Nominal by Nominal	Phi	.184	.001
	Cramer's V	.106	.001
N of Valid Cases		970

a Not assuming the null hypothesis.

b Using the asymptotic standard error assuming the null hypothesis.

Interpreting the Output:

The first table is our crosstab which you are expert at interpreting by now. The second table is our test for statistical significance, which you also know how to interpret. The last table is our strength of association measures. Although the mathematics behind these measures vary, the results are often very similar. The second column tells you the name of the measure of association. The Value column shows the strength of the relationship. Since both measures are not all that far from 0, this is a weak relationship. The next column tells you the statistical significance. In this analysis, the results are statistically significant.

Exercise 22:

Is there a relationship between gender and views on spanking? Is that relationship statistically significant?

Is there a relationship between religious preference and spanking? Is that relationship statistically significant?

2. Measuring Relationships for Ordinal Data:

Tau Tau B is considered to be a conservative function because it subtracts ties, therefore the results tend to be smaller. A tau b of .2 or more is respectable.

Gamma: Gamma is based upon probability. If you select on individual at random it computes what the chances of getting the same answer if you selected another at random. A gamma of .3 or more is worth examining.

Spearman=s Rho: Converts interval data to data which can be ranked (highest to lowest)

Using SPSS: Output for Relationships Using Ordinal Data

Exercise 23: What is the strength of the relationship between political ideology and legalization of marijuana?

Steps:

Click Analyze => Descriptive Statistics => Crosstabs

Put Political ideology (politics) in row and legalization in column

Click on cells to do your percentage on row.
Click continue.
Click on the statistics at the bottom of this box.
Go to Ordinal, and select tau and gamma.
Click OK.

You will get a normal crosstabs table like the one for exercise 16. At the bottom, the computer will generate additional tables with the measures of association.

Output 23:

Symmetric Measures

		Value	Asymp. Std. Error	Approx. T	Approx. Sig.
Ordinal by Ordinal	Kendall's tau-b	.210	.029	6.917	.000
	Kendall's tau-c	.220	.032	6.917	.000
	Gamma	.391	.052	6.917	.000
N of Valid Cases		899

Using SPSS to Do Interval Level Correlations:

The most common test to determine relationships between interval level variables is Pearson=s R. The coefficient compares the data set to the ideal perfect data set and to no relationship. Then it assigns a score (0-1) depending on how close the data is to the ideal set.

Exercise 24: What is the relationship education and income? Our hypothesis is that those with more education would earn more. Both years of education and income are at least interval level data.

Steps:

Click on Analyze

Click Correlate,
Click on Bi-Variate. A new screen will pop-up and you will see you list of variables along the left side.
Select IncomDol (income recoded to dollars) and move to box on right.
Select Educ (highest year of education) and move to box on right.
At the bottom of window, select Pearson=s r

Click OK

Output 24:

Correlations

		EDUC Highest Year of School Completed	INCOMDOL Income recoded to dollars
EDUC Highest Year of School Completed	Pearson Correlation	1.000	.460
	Sig. (2-tailed)	.	.000
	N	1496	1362
INCOMDOL Income recoded to dollars	Pearson Correlation	.460	1.000
	Sig. (2-tailed)	.000	.
	N	1362	1365

** Correlation is significant at the 0.01 level (2-tailed).

Regression Analysis:

When doing regression analysis you will be working with an equation. This equation will allow you to describe a data set, to estimate population parameters, to infer causality, and to forecast.

Measures of Association: R

r= multiple correlation coefficient (overall fit).

r² = proportion of explained variation

1-r² = proportion of unexplained variation (a.k.a. coefficient of determination).

Using SPSS to Do a Simple Regression

Exercise 25: Let=s look at education and income, using a simple regression

Steps:

Click on Analyze
Click on Regression
Click on Linear.
Select your variables:

Income (Incomdol) and place in dependent box

Educ (highest year of education in independent box

Click OK.

Output 25:

Model Summary

Model	R	R Square	Adjusted R Square	Std. Error of the Estimate
1	.460	.212	.211	22699.57

a Predictors: (Constant), EDUC Highest Year of School Completed

ANOVA

Model		Sum of Squares	df	Mean Square	F	Sig.
1	Regression	188426936995.327	1	188426936995.327	365.685	.000
	Residual	700768148035.877	1360	515270697.085
	Total	889195085031.204	1361

a Predictors: (Constant), EDUC Highest Year of School Completed

b Dependent Variable: INCOMDOL Income recoded to dollars

Coefficients

		Unstandardized Coefficients		Standardized Coefficients	t	Sig.
Model		B	Std. Error	Beta
1	(Constant)	-13936.587	2671.534		-5.217	.000
	EDUC Highest Year of School Completed	3802.543	198.848	.460	19.123	.000

a Dependent Variable: INCOMDOL Income recoded to dollars

Interpreting the Output:

We have three tables. The first table, The Model Summary, shows us how well a model explains variation in income. The R square is .21, which indicates a moderate relationship: 21% of all the variation in income is explained by education. The second table, ANOVA, tells us the statistical significance of the model. We want to look at the last column which gives us the statistical significance. In this case, it is less than .05, so the results are statistically significant. The coefficient table shows the impact of education on income. For every year of education, income goes up $3,802. This is also statistically significant.

Using SPSS to Do Multiple Regression:

To determine the regression coefficients and the constant you follow the same steps as if you are doing a bi-variate analysis. Once you have entered you dependant variable you can simply enter more than one independent variable. The computer will generate for you the coefficients the constant and the R value.

Exercise 26: What explains income?

Steps:

Click Analyze

Click Correlate

Click Linear

Select IncomDol and move to dependent box
Select EDUC (highest year of school complete, AgeWEd, and DWELOWN (Homeowner or renter) and move to the Independent Box
Click OK

Output 26:

Model Summary

Model	R	R Square	Adjusted R Square	Std. Error of the Estimate
1	.577	.333	.330	21048.82

a Predictors: (Constant), DWELOWN Homeowner or Renter, AGEWED Age When First Married, EDUC Highest Year of School Completed

ANOVA

Model		Sum of Squares	df	Mean Square	F	Sig.
1	Regression	160736543078.188	3	53578847692.729	120.931	.000
	Residual	321656336117.017	726	443052804.569
	Total	482392879195.205	729

a Predictors: (Constant), DWELOWN Homeowner or Renter, AGEWED Age When First Married, EDUC Highest Year of School Completed

b Dependent Variable: INCOMDOL Income recoded to dollars

Coefficients

		Unstandardized Coefficients		Standardized Coefficients	t	Sig.
Model		B	Std. Error	Beta
	(Constant)	7174.054	5114.637		1.403	.161
	EDUC Highest Year of School Completed	3880.293	255.097	.481	15.211	.000
	AGEWED Age When First Married	-142.368	161.734	-.027	-.880	.379
	DWELOWN Homeowner or Renter	-13176.228	1601.745	-.253	-8.226	.000

a Dependent Variable: INCOMDOL Income recoded to dollars

Interpreting the Output:

How good is our model?

The Model Summary gives us the answer here in the R square value. This tells us how much variation in income is explained by this model. Overall, this model explains .33 of the variation, which indicates there is a moderately strong relationship here. (Note: this is higher than what we got from our simple regression, so we are on the right track here).

Is our overall model statistically significant?

The ANOVA table gives us a lot of information. The only thing of importance for us is the last column, which gives us the significance. In this case, it is less than .05, so the model is statistically significant.

What explains income?

The Coefficient Table: The unstandardized coefficients tell you how much change there will be in income for every unit change in each independent variable while holding the others constant. (Sometimes we talk about this as the independent effect of each of the variables in our equation on income). So if we look at the independent effect of education while holding age when married (age wed) and home ownership (dwelown) constant, for every year increase in years of education, we get a $3,880 increase in income. Looking at home ownership, we see a negative sign, indicating an inverse relationship. In this case, we need to know how this variable was coded (right click on the variable to get information when you are in the regression window). This variable was coded 0 if the respondent owned a home and 1 if respondent rented. This gets interpreted as we move from a home owner (0) to a renter (1), income goes down by $13,176. Because of the way the data was coded, this is an inverse relationship.

Standardized Coefficients: Beta weights: The measurement of our independent variables are not the same; we use years for education and homeowner/renter for dwelling. They are not comparable. So, the computer standardizes them so we can see which is stronger in changing the dependent variable. It is called the Standardized Coefficient (or beta or beta weights). In this case, Education has a higher Beta weight.

Other Stuff:

Managing the Output:

Deleting stuff from an Output file:

Option A: right click on the chart or table, then hit delete button.

Option B: go to the file menu on the left side of your screen. Click on the yellow file folder, and it will delete everything in that folder when you hit the delete button. Or you can click on the specific files and hit delete button.

Print an Output File

Click File C> Print

You can print the material you have selected and highlighted by clicking on selection

Or You can print the whole file (remember to deleted stuff you don=t want) by clicking All Visible Output

Helpful Hint: You can save time and ink by clicking on properties and selecting econofast print.

If you want to view the output before printing. File==>Print PreView
If you want to change the page layout, click on file==>Page Setup

You can choose between Portrait (which this is) or Landscape.

Crosstabs print better using landscape.

Note: all your output gets saved to a single file until you close it (either save or delete). If you hit print, all the output done in the output session will be printed out.

Saving an Output file:

go to output file. Click on FileB>Save As

Select where you want to save it: A drive at the school, what drive you want at home
Give it a short name: Class1 or Output1
Click Save.

I don=t usually save output since it is easy enough to recreate. But you may want to save the output from assignments at least until after your receive feedback.

Managing the Hardcopy Output:

I tend to save the hardcopy output while I am working on a report because it maintains an audit trail as GAO folks refer to it. I label output and keep it in a looseleaf binder. I label it in a way I can find it when I write a draft report using data and I can source it (e.g. Table 1 from Run 1, pages 12-20). It makes it easier to find the data if there is a question about the analysis or if I want to have someone check that my numbers matchBaccuracy check.

You can label SPSS output before printing or saving.

Exercise: For example, you want to label the run that has information about demographics:

File ==>open==>students.sav

analyze=>descriptive statistics==>frequencies

Select demographic variables (sex, race, etc) and click => to move them to analysis box
Click OK

The output will pop up.

Double Click on Frequencies and a box will appear.
You can now edit the box
You just type in what you want, eg. Demographic, Run #, Date, Your Name
Then move the cursor outside the box and click. It is changed.

You can change the label of any table as well. Double clicking creates an edit box.

Moving a SPSS Table into a Word Document:

You may find it helpful to move SPSS tables into your report. You want to open a word document. You can do this my hitting the B button at the top of the SPSS menu, then open up word and open a blank file. Got it? Then minimize that and go back to SPSS.

Let=s say you want to move the table reporting sex to your word document.

Steps:

In the output, double click on the Asex@ table.
Right click, and select copy.
Open up your word document, place the cursor where you want the table to go, right click and select paste.
I do not like the SPSS tables all that muchBthey are too technical and ugly, so you may want to clean them up. You can change the table in word using the Table options.
Go into the table and right click. Options will appear.
Select speed format. Select a format that you like. I often use single lines.
You can delete the 1^st column, and widen the columns so the labels re not cut off.

Creating Summary Tables: You cannot create a single table that shows the percent rating various instructional approaches as effective in SPSS. You will have to create the table in word or wordperfect, and type in the percent distributions.

Using Graphs to Analyze and Display Data

This section will walk you through using graphs options. Remember to use Bar or Pie charts for ordinal and nominal variables.

Exercises: Show a picture of the percent of respondents in each program.

Steps:

( A )Bar Charts:

Click Graphs from the menu bar
Click Bar.
In the >Bar Charts= dialogue box click on simple
Select >data chart for groups of cases@

Select >% of cases= in >Bars represent= box
Click define.
Select program variable and move it to category axis

Click on >Options= button and remove check mark in Display groups defined by missing values. You do not want missing values in your charts as a general rule..
Click Continue.
Click the Title button and add a title if you like to have one.
Click OK.

Editing the chart: You edit the chart in the output box. Double click on the chart. This will pop up the chart editor. You can change the fill, the color, the lines, bar style, bar labels, and the direction (from vertical to horizontal). You need to click apply in the little pop boxes for each of these changes. When done, hit the X button. When done with the editor, you click the X button.

You can move the chart to a word document. In the Output, double click on the chart. It will come up as box. Right click, and select copy as object.or copy: I am not sure how much difference there is. But play with thisBthe differences may become clear.

( B )Pie Charts:

Click --> Graphs -->Pie.
Select >summaries for groups of cases=,
Click define.
Select >Sex= variable and move it to >Define slices by= box.
Click on >Options= button and unmark >Display group defined by missing values= to get a chart only with Avalid@ responses
Click continue.

Click on the >Title= button and add a title, footnote
Select >% of cases= in >Slices represent= box
Click >OK=.

Editing the chart: Go to editor by double click. You can pull out a slice, by clicking on that slice and then selecting the button that looks like pac man; if you don=t like it, click pac man again. You can change color, fill and lines.

You can add percents to the slices. While still in the chart editor box, select Chart from the menu at top of editor box, then click options. In options, and select percent (clicking on it gives it check mark). Click on format, and change decimal places to 0, then click OK to get out of the options box..

( C ) Cluster Charts:

Click --> Graphs -->Bar-->Clustered.
Select >summaries for groups of cases=

In the Bars represent box, select % of cases.
Click define.
Select >teach5= variable and move it to Category Axis box.
Select >Program= variable and move it to Define Clusters by box.
Click on >Options= button and remove the check mark by clicking on Display group defined by missing values, so missing values won=t ge in the chart.
Click Continue.

Click on the >Title= button and add a title, footnotes, whatever.
Click >OK=.

Note: These are the same variables as used in crosstabs. This is another way of describing those variables with charts option. Once you are familiar with the creation of charts to represent your data analysis, you can explore different options provided by SPSS application (like - line charts, area charts, stacked bar charts, scatter charts and boxplots etc., )to better describe your data graphically.

Again, you can go to the chart editor (double click on chart) and make whatever changes you would like. Remember, color does not always reproduce well in black and white, so you might find using different fill to be more effective.