SPSS

Basic Concepts

Levels of measurement | Significance
- YouTube video reference
- Significance -- Confidence Level
  - 95%, p < .05, t-value > 1.960 (df = infinity)
  - 99%, p < .01, t-value > 2.576 (df = infinity)
  - 99.9%, p < .001, t-value > 3.291 (df = infinity)
- Significance -- Confidence Interval
  - Lower, Upper
- Defining variables in SPSS - - - Video 6min10sec
- How to Import data from an excel file to SPSS
- How to calculate mean scores in SPSS
- How to transform and reverse scores (recode) in SPSS

Significance

Confidence Level

95%, p < .05, t-value > 1.960 (df = infinity)
99%, p < .01, t-value > 2.576 (df = infinity)
99.9%, p < .001, t-value > 3.291 (df = infinity)

Confidence Interval

Lower, Upper

top

Defining variables in SPSS

Variable view

Variable Name: Not more than 8 characters, not start with a number, no space or symbol

Label: Any number of character, can have space or symbol

Type: Numeric for analysis

Variable Type

String: including text or some special character, need to clean up and change back to Numeric for calculation

Decimals: Change to zero if it is an integer

Value

Value Labels:

7-point, 5-point Likert's scale
Add 1-7 with value labels

Missing

Missing Value:

Can just leave it blank
Just to make sure you have input the data
Sometimes, you forget whether you have input or not
By practice, type 99, 999 or 9999 (depending on whether there is the same number/value in the valuable

Measure: Scales of measurement

Nominal: By category
Ordinal: By ranking of order
Scale: Interval and ratio, continuous variable, for calculation of means and standard deviations

Data view

top

How to Import data from an excel file to SPSS

Import data from an excel file to SPSS

Steps: File >> Import Data >> Excel

Read variable names from first row of data

variable names should not have more than 8 characters, no space, start with an alphabet, not with a number

Type: Numeric

Imported data may wrongly be interpreted as "String"
Change back to "Numeric" for calculation

Decimal: Change to "0" if decimal places are not needed

Value

Value Labels

Name each score

For example, for a 7-point Likert Scale

"1" - Strongly disagree
"2" - Disagree
"3" - Somewhat disagree
"4" - Neutral
"5" - Somewhat agree
"6" - Agree
"7" - Strongly agree

Value:

Copy Value and paste to all

Missing

Missing Values:

Type the missing value so that SPSS will not wrongly calculate the missing data, e.g., 9, 99 or 999 (Make sure not to overlap with the data)

Measure

Scales of Measurement

Imported data may wrongly be interpreted as "Nominal" even they are continuous numbers
Change "Measure" to "Scale

top

How to calculate mean scores in SPSS

Steps: Transform >> Compute Variable

Target Variable: Create a new variable and give a name
Numeric Expression: Add all the indicators/items in the variable/construct, divided by the total number of items
E.g., POAM = (POAM1 + POAM2 + POAM3 + POAM4 + POAM5) / 5

OR Select a formula for the calculation

Function group: Select a category from the list, e.g., All
Functions and special variables: Select a formula from the list, e.g., Mean

E.g., PORC = (PORC3 + PORC4 + PORC5) / 3

E.g., OKSB = (OKSB1 + OKSB2 + OKSB3 + OKSB4 + OKSB5) / 5

Descriptive Statistics:

To report the descriptive statistics for the newly created variable (summed/average score of each variable)
Steps: Analyze >> Descriptive Statistics >> Descriptives

Select only the summed score variables
E.g., POAM, PORC, OKSB

Results: Descriptive Statistics Table, including minimum, maximum, mean, standard deviation

Regression: To test the relationship between the variables

Steps: Analyze >> Regression >> Linear

Dependent variable: OKSB
Independent variables: POAM, PORC

Regression Results

Model Summary: R Square
ANOVA: F-value and Significance
Coefficients: Standardized coefficients beta, t-value, sig.

top

How to transform and reverse scores (recode) in SPSS

Record each old value to a new value

Steps: Transform >> Record into Different Variables

From the left-hand side, select the variable to record, click the middle arrow and move the variable to the right-hand side

In the Output variable:
- Name: Give a new name
- Label: Give a new label
- Click "Change" to update the output variable
Click "Old and New Values"

Old Value: Click "Value" and input a value
New Value: Click "Value" and input the changed value

For example,

1, 2, 3 are recoded to 1 (Range, lowest through value, input "3")
4 is recorded to 2 (Old Value, "4"; New Value, "2"
5, 6, 7 are recorded to 3 (Range, value through highest, input "5")

Reverse values for a negation item

Steps: Transform >> Compute Variable
Target Variable: Input a new variable name, e.g., OKSB1ve
Numeric Expression: 8 - OKSB1 (reverse a 7-point Likert Scale)
- 8-"7" = "1"; 8-"6" = "2"; 8-"5" = "3"; 8-"4" = "4"; 8-"3" = "5"; 8-"2" = "6"; 8-"1" = "7"

top

Descriptive analysis

How to run frequency distribution in SPSS - - - Video 4min13sec
How to run descriptive analysis in SPSS

How to run frequency distribution in SPSS

Step:

Analyze >> Descriptive Statistics >> Frequencies

Select the variable on the left-hand side

Click the middle arrow to move the variable to the right-hand side, e.g., gender

Select "Charts" if you need to draw a chart

Select a chart type, e.g., "Bar charts"

Click "Continue" and then click "OK"

The results show the Frequency Table, with frequency, per cent, valid per cent and cumulative per cent

The results show the Bar Chart

top

How to run descriptive analysis in SPSS

Step: Analyze >> Descriptive Statistics >> Descriptives

Click to select a variable from the left hand side, e.g., score

Click the middle arrow to move it to the right hand side

Click "Options" for more

Confirm mean, standard deviation, minimum, maximum are selected

Click "Continue", then click "OK"

Results show the Descriptive Statistics Table, including the variable name, the number of the sample, minimum, maximum, mean and standard deviation

top

Means Tests

How to run one-sample t-test in SPSS - - - Video 5min33sec
How to run two independent samples t-test in SPSS - - - Video 6min56sec
How to run paired-samples t-test in SPSS - - - Video 7min40sec
How to run one-way ANOVA and post hoc analysis in SPSS - - - Video7min51sec

How to run one-sample t-test in SPSS

How to run one-sample t-test

To compare the mean of the sample to the mean of the population.

The measurement of the variable should be continuous (scale), e.g., IQ score.

Step: Analyze >> Compare Means >> One-Sample T Test

Click to select the variable from the left-hand side

Click the middle arrow to move to the right-hand side

Type the mean of the population for comparison to the "Test Value", e.g., 105

Check "Option" for Confidence Interval Percentage, e.g., 95%

Results show the descriptive statistics table, including the sample size, mean, standard deviation and standard error mean

Results show the One-Sample Test, including the t-value, degree of freedom (df), significance (2-tailed)(p-value), mean difference, the lower and upper value of the 95% confidence interval of the difference

How to write the report

top

How to run two independent samples t-test in SPSS

How to run two independent samples t-test

The data looks something like this.

There are two conditions, 1 and 2.
For example, A class is teaching face-to-face where B class is teaching online.
For example, Group A is taking a placebo where Group B is taking a new medicine.

Steps: Analyze >> Compare Means >> Independent-Samples T Test

Test Variable(s)

Click and select the left-hand side variable for comparing means, e.g., score
Click the middle arrow to move to the "Test Variable(s)"

Grouping Variable

Click and select the left hand Grouing Variable, e.g., condition
Click the middle arrow to move to the "Grouping Variable"
Click "Define Groups"

The value in the condition is 1 and 2, therefore, put 1 in Group 1 and 2 in Group 2

Click "Continue" and click "OK"

Results show the descriptive statistics, including the condition 1 and 2, sample sizes, means, standard deviations, standard error means

Results show the Independent Samples Test, including under the "Equal variance assumed," the t-value, df, sig., mean difference, standard error difference, the lower and the upper value of the 95% confidence interval of the difference (the mean difference can be any value within this range)

How to write the report

top

How to run paired-samples t-test in SPSS

How to run paired-sample t-test

To compare the means, the variable should be continuous (Scale)

The data file should look like this:

There are two measurements of the same subject
For example, before the treatment (pre) and after the treatment (post)
For example, first taught by face-to-face, then taught by online

Steps: Analyze >> Compare Means >> Paired-Samples T Test

Click and select the left hand side variables

Click the middle arrow to move to right hand side "Paired Variables" one by one, Variable 1 and Variable 2

Results show the descriptive statistics, including the two measurements, the means, the sample size (Must be the same), standard deviations, standard error means

Results show the paired samples correlations

As the same subject was measured two times, there existed correlation between the two measurements

Results show the Paired Samples Test, including the mean difference between the two measurements, the standard deviation, the standard error mean, the lower and the upper value of the 95% confidence interval difference, the t-value, df (n-1), sig.

If you would like the post value to minus the pre value, place the [post] variable to Variable 1, the [pre] value to Variable 2

Paired Samples Statistics are the same

Paired Samples Correlations are the same

The mean difference will be positive in the Paired Samples Test

How to write the report

top

How to run one-way ANOVA and post hoc analysis in SPSS

How to run one-way ANOVA

One-way ANOVA compares means of 3 or more groups

One variable is the grouping variable
One variable is the measurement for comparison

The data file looks like this

Grouping variable with values 1, 2, and 3 (3 conditions)
Measurement for comparison should be a continuous variable (Scale)

Steps: Analyze >> Compare Means >> One-Way ANOVA

Select from the left-hand side variables and click the middle arrow to move to the right-hand side

Dependent List: The measurement variable
Factor: Grouping variable

Post Hoc analysis

One-way ANOVA provides only significant or not significant for the means but will not provide more information
Post Hoc analysis further compares the means between each pair of groups

Steps:

Click "Post Hoc" analysis
Select from the "Equal Variances Assumed" list, for "Tukey"

Results show the ANOVA table if the means are significantly different, e.g., Sig. (p-value) < .05

Post Hoc Tests

Compares each pair of groups for the mean difference

For example,

Group 1 and Group 2, mean difference is 6.70 (p < .001)
Group 1 and Group 3, mean difference is 5.10 (p < .001)
Group 2 and Group 3, mean difference is -1.60 ( p = .113, not significant)

Conclusion

Group 2 and Group 3 are a homogeneous subset which has a significant mean difference with the Group 1 subset

Means

ANOVA analysis does not provide the means
Steps: Analysis >> Compare Means >> Means

From the left-hand side variables to select and to click the middle arrow to move to the right-hand side

Dependent List: measurement variable
Independent List: grouping variable

Results show the Means Report table for each group, including means, sample sizes, standard deviations

How to write the report

top

Testing for Relationship

Correlation

How to run correlation in SPSS - - - Video 2min10sec

Regression

How to run simple regression in SPSS
How to run multiple regression in SPSS - - - Simple & Multiple Linear Regression Video 9min30sec

Chi-square

How to run chi-square test in SPSS - - - Video 3min57sec

How to run correlation in SPSS

Correlation analysis aims to find if variables are significantly related, either positively or negatively related (inverse relationship)

Variables should be continuous (Scale) for calculation

The data looks like this

Steps: Analyze >> Correlate >> Bivariate

From the left hand side, select the variables and click the middle arrow to move them to the right hand side (Variables)

Correlation Coefficients: Pearson
Test of Significance: Two-tailed

Results show the Correlations Table

To report the results, the correlations table needs to be cleaned up

To display only the lower triangle of a correlation matrix
Right click to select "Export"

Save Document Type as, "Excel 97-2004 (*.xls)"

Select the location and give a file name

Open the excel file

Remove all the unnecessary columns and rows

Leave only the variable and the correlation coefficients

Finally, present the Correlation Matrix as below

To display only the lower triangle of a correlation matrix
Adjust the decimal places to display consistently the correlation coefficients, e.g., 2 decimal places for all

top

How to run simple regression in SPSS

Simple Regression:

An independent variable is hypothesized to be significantly related (either positive or negative) to a dependent variable
Both variables are continuous (Scale)

The data looks like this

Steps: Analyze >> Regression >> Linear

From the left-hand side, select the variable and click the middle arrow to move to the right-hand side:

Dependent: Place the dependent variable
Independent: Place the independent variable
Method: Enter

Results: Variables Entered/Removed

List the Variable Entered: Independent variable
List the method use: Enter
Note the dependent variable

Results: Model Summary

R
R-Square
Adjusted R Square
Standard Error of the Estimate

Results: ANOVA table

Regression and Residual
Sum of Squares, df, Mean Square, F-value, Sig.

Results: Coefficients

Constant and Independent Variable
Unstandardized Coefficients (B, Standard Error)
Standardized Coefficients (Beta)
t-value, Sig.

top

How to run multiple regression in SPSS

How to run multiple regression

Multiple Regression:

There are several independent variables which are hypothesized to be significantly related (either positive or negative) to a dependent variable (SPPS accepts only one dependent variable)
All variables are continuous (Scale)

The data looks like this

Steps: Analyze >> Regression >> Linear

From the left-hand side, select the variable and click the middle arrow to move to the right-hand side:

Dependent: Place the dependent variable
Independent: Place all the independent variables

Method:

Enter
Stepwise
Remove, Backward, Forward

Results: Variables Entered/Removed

List the Variable Entered: All the independent variables
List the method use: Enter
Note the dependent variable

Results: Model Summary

R, R-Square, Adjusted R Square, Standard Error of the Estimate

Results: ANOVA table

Regression and Residual
Sum of Squares, df, Mean Square, F-value, Sig.

Results: Coefficients

Constant and Independent Variables
Unstandardized Coefficients (B, Standard Error)
Standardized Coefficients (Beta)
t-values, Sig.

How to write the report

top

How to run chi-square test in SPSS

Chi-Square Test

To find the relationship between two variables
Measurement of the variables are categorical or discrete, not continuous

Data looks like this

Steps: Analyze >> Descriptive Statistics >> Crosstabs

By practice

Row(s): for Independent variable
Column(s): for Dependent variable

Option, "Cells":

Accept the default value: Observed

Results: X*Y Crosstabulation

Option, "Cells":

Checked more values:
Counts: Observed, Expected
Percentages: Row, Column, Total

Results: X*Y Crosstabulation

Provide more details

top

Construct Validation

(For construct having more than one item) --- A Complete Study Video 27min13sec

Descriptive analysis for means and standard deviations of variable items
Construct Validation
- Reliability or Internal Consistence (for each variable)
  - Cronbach's alpha > .7
- Construct validity (Factor Reduction, Factor loadings, principal component analysis, varimax rotation, eigen values, % of variance explained)
  - Exhibit Convergent validity (factor loadings > .7)
  - Exhibit Discriminant validity (No crossloading or not serious below .3/.4)
Compute summed average score for each variable
Descriptive statistics for summed average score
Multiple Linear Regression
- Model summary, R-square
- ANOVA, F-value, sig,/p-value
- Standardized coefficient beta, p-value

Reliability / Internal consistency
- How to analyze internal consistency in SPSS
Factor analysis (discriminant/convergent validity)
- How to run factor reduction in SPSS

How to analyze internal consistency in SPSS

How to calculate Reliability (Internal Consistency)

Steps:

Analyze >> Scale >> Reliability Analysis

Select all the items in ONE construct

Click the middle arrow to move to right hand side "Items"
Model: Alpha

Click "Statistics"

Descriptives for

Check "Item"
Check "Scale"
Check "Scale if item deleted"

Results: Case Processing Summary

Sample size

Reliability (Internal Consistency) Statistics:

Cronbach's Alpha, literature recommended > 0.7
e.g., 0.903 > 0.7 for all the 5 items of POAM
POAM is internally consistent

Item Statistics

mean, standard deviation, sample size

Item-Total Statistics

Cronbach's Alpha if Item Deleted: Check if remove any one of the items, the Cronbach's Alpha value can be uplifted, then remove the item
Especially, sometimes, an item deviates a lot from the other items. Remove it can increase the alpha value a lot
Removal of any one item of POAM will get the alpha value between 0.869 and 0.897
Not more than 0.903, all 5 items of POAM
Therefore, no item will need to be removed from POAM

Scale Statistics

man, variance, standard deviation, number of items

Repeat the steps for another construct

PORC: 5 items

Results: Reliability (Internal Consistency) Statistics:

Cronbach's Alpha, literature recommended > 0.7
e.g., 0.908 > 0.7, for all the 5 items of PORC
PORC is internally consistent

Results: Item-Total Statistics

Removal of any one item of PORC will get the alpha value between 0.875 and 0.896
Not more than 0.908, all 5 items of PORC
Therefore, no item will need to be removed from PORC

Repeat the steps for another construct

OKSB: 5 items

Results: Reliability (Internal Consistency) Statistics:

OKSB Cronbach's Alpha = 0.923 > 0.7 for all the 5 items
OKSB is internally consistent

Item-Total Statistics

Removal of any one item of OKSB will get the alpha value between 0.896 and 0.916
Not more than 0.923, all 5 items of OKSB
Therefore, no item will need to be removed from OKSB

To report reliability in a descriptive statistics table

Steps:

Analyze >> Descriptive Statistics >> Descriptives

Select all items from the left-hand side

click the middle arrow to move all to the right-hand side, "Variable(s)"

Results: Descriptive Statistics Table

List of each item, sample size, min, max, mean, standard deviation

Right click >> Copy As >> Excel Worksheet

Open the Descriptive Statistics excel file

Add a column, "Cronbach's Alpha"
Add the Cronbach's Alpha value for each construct
POAM: 0.903; PORC: 0.908; OKSB: 0.923

top

How to run factor reduction in SPSS

Steps:

Analyze >> Dimension Reduction >> Factor

Move all the items to the Variables

click Extraction

Method: Principal components

click Continue
clcik Rotation

Method: Varimax

Therefore, the items will be grouped to corresponding factors (components) accordingly
click Continue

Scores

click Continue
click Options

Coefficient Display Format

check, Suppress small coefficients
Absolute value below: 0.4
- If all the factor scores are displayed, it is not easy to read
- If factor scores below 0.4 are not shown, it is easier to identify the serious cross-loadings
- Items with cross-loadings (loaded to more than one factor/component) should be removed, they cannot exhibit discriminant validity

Results: Communalities

The list of extraction values for each item

Results: Total Variance Explained

Read for components where Eigenvalues > 1
Check the total variance explained (% of variance, cumulative %)

Results: Principal Component Matrix

Not so useful as the components are not distinct

Results: Rotated Component Matrix

This table is for report
Each item is loaded clearly to the corresponding component
Convergent validity: The factor loadings are strong: (1) They are acceptable if the value is > 0.5; (2) They are significant if the value is > 0.7
Discriminant validity: The factors are distinct, i.e., there is no cross-loading where an item will not belong to more than one factor/component