SPSS video lectures

Question 1

What is skewness?

Answer 1

tail leans to left or right

Question 2

How can you get an indication of the normality (normal distribution) on the data depending on what?

Answer 2

By looking at the mean and the median. The closer they are, the more likely it is that its normally distributed. its just a relative question.

Question 3

What is kurtosis?

Answer 3

data is peaked or flat. we want a bell-shaped curve

Question 4

What is the rule of thumb for skewness and kurtosis?

Answer 4

an absolute value of one

Question 5

If you have a test of normality and you have a sample size above 50, what do you use?

Answer 5

the Kolmogorov-Smirnov test

Question 6

If you have a test of normality and you have a sample size below 50, what do you use?

Answer 6

the Shaoiro-Wilk

Question 7

When having a test of normality, what is the rule of thumb for the significance?

Answer 7

It needs to be above 0.05. Because a value under 0.05 indicates non-normality. (abnormal distribution).

Question 8

How can you interpret a box plot?

Answer 8

The more symmetric the boxplot, the more normally distributed the data is.

Question 9

Correlation varies from what interval?

Answer 9

-1 (perfectly negatively correlated) to +1 (perfectly correlated

0 is no correlation

if its below 0.3 don’t consider it
0.3 -> 0.6 medium sized correlation
above 0.6 strong correlation

Question 10

What does it mean when we say that the correlation coefficient is significantly different from zero?

Answer 10

the correlation coefficient is significant

Question 11

What is independent samples t-test? Give an example

Answer 11

Independent samples t-test is when we have to group means (like men vs women average income) that we want to compare to see if the means are significantly different from each other, the same or just random chance. in this case the two groups are not related to each other.

not to be compared wit repeated measure t-test: womens income at time 1 and then time 2. two dependent group (same group at different times)

Question 12

When we have dummy variables, always do what when coding?

Answer 12

code them as 0 and 1.

Question 13

What is Levenes test for equality of variances?

Answer 13

this is one of the assumptions of the independent samples t-test. the two groups should have equal variances. if we have quite different group sizes, there is a chance the variance will not be equal.

Question 14

What is the standard cutoff for significance?

Answer 14

0.05

> 0.05 = equal variances

< 0.05 = would be significant difference

Question 15

What is the rule of thumb for t-tests?

Answer 15

if the t value < 2, then its prob not significant

Question 16

What are the rule of thumbs for Cohens D point estimate?

Answer 16

anything around 0.2 would be a small effect
< 0.2, dont interpret it
0.5 medium effect
0.8 large effect

Question 17

What is paired samples/repeated measures t-tests?

Answer 17

you look at the same period but for different time periods. they are dependent on each other.

Question 18

What does the significance level in a paired sampled t-test tell us?

Answer 18

if its < 0.05 we can say that there is a significant difference between the means
if its larger than > 0.05 the difference are by chance, there is no significant difference

Question 19

What is Exploratory factor analysis?

Answer 19

It is a way of establishing construct validity which is made up of two parts: convergent validity and discriminant validity.

Question 20

What is a latent construct? (EFA)

Answer 20

Something that is not directly observable, like collaboration or trust so we observe through ex a questionnaire. Everything in like colllaboration should converge. But other factors like trust should diverge from each other. (discriminant). because otherwise trust and collaboration is overlapping.

Question 21

Is EFA subjective or objective?

Answer 21

subjective

Question 22

A factor coefficient can be anywhere from… what range?

Answer 22

-1 to +1

0.4-0.5 (absolute value) is meaningful

being close 0 is meaningless

Question 23

How can you know if the factor ability of the data is good?

Answer 23

Look at the correlation matrix. But we also have the KMO and Bartletts test

Question 24

What is Kaiser-Meyer-Olkin measure (KMO) of sampling adequacy?

Answer 24

The Kaiser-Meyer-Olkin (KMO) measure is a statistic that helps you decide if your data is suitable for factor analysis, which is a method used to identify patterns or groupings in data.

Key Points in Simple Terms:
1. Purpose: It checks if the variables in your dataset are related enough to each other to form groups (factors).

2. Scale:
   
   • The KMO value ranges from 0 to 1.
   • Higher values (closer to 1) mean that your data is good for factor analysis.
   • Lower values (closer to 0) mean that your data is not suitable for factor analysis because the variables don’t relate well to each other.
3. Rule of Thumb:
   
   • KMO ≥ 0.8: Excellent.
   • 0.7 ≤ KMO < 0.8: Good.
   • 0.6 ≤ KMO < 0.7: Acceptable.
   • KMO < 0.6: Poor, factor analysis might not work well.
4. How It Works: The KMO compares:
   
   • Correlations: How strongly variables are related.
   • Partial Correlations: How much the relationship between two variables is explained by other variables.
     If correlations are much stronger than partial correlations, the KMO will be high, meaning factor analysis is appropriate.

Example:
If you have a survey with several questions, the KMO helps you know if the questions are related enough to summarize them into fewer factors (like grouping questions about “job satisfaction” together).

In summary, the KMO is like a “test of fit” to see if your data can be simplified and grouped through factor analysis!

the absolute lowest value for KMO is 0.5

0.9 is close to 1 so its Marvelous. you prefer a higher KMO

Question 25

What is the Anti-image matrix?

Answer 25

lets say that our KMO is 0.5 or 0.4. its low and we are not happy with it. what we do is look at the anti-image correlation and we look at the diagonal. what could happen is that lets its a bad indicator; that would have great effect on the KMO. so if we remove that from the analysis our KMO would go up. so the anti-image corr matrix shows our each contributor to the KMO. so we can identify one indicator or two that we can remove.

Question 26

What are communalities?

Question 27

What is the rule of thumb for communalities?

Answer 27

We want them to be above 0.5. Ex, we have 4 that is 0.4, the other are well above 0.5. EFA is subjective and we dont go through it using absoulte, we go through it collecting evidence. So our KMO was fine but here we get an indication that some factors perhaps dont share enough common variance to remain in the model. But we would not remove them. We just know some of them would be problematic.
the lower the communality, the more problematic.

Question 28

What about the total variance explained table in EFA?

Answer 28

Here we want to look at the cumulative variance explained for the solution. And how many factors were extracted? Did we want 3? Ok, is it reasonable given how close to 1 the Eigenvalue is? You can force a 3 factor solution

Question 29

What can you do to make a Rotated Factor Analysis easier to read?

Answer 29

Suppress small values (below 0.4-0.5).

Question 30

When you have a 3 factor solution, how can you know in percentage how much of the variance is explained in the factor model?

Answer 30

the cumulative % under initial eigenvalues

Question 31

When do we know that we have discriminant validity?

Answer 31

When each factor indicator loads above 0.5 on one factor and below 0.5 on the other factors

Question 32

We have convergent validity in a rotated factor matrix when…

Answer 32

each indicator is loaded on the proper factor (ensamma på sina ställen)

Question 33

We have discriminant validity in a rotated factor matrix when…

Answer 33

they are not loading on the wrong places.

Question 34

What is another word for Reliability analysis?

Answer 34

Cronbachs alpha

Question 35

What is the generally accepted cutoff value for Chronbachs alpha?

Answer 35

above 0.7 gives a reliable construct

Question 36

What is another term for multiple regression model?

Answer 36

ordinary least squares regression

Question 37

What is another word for R squared?

Answer 37

explained variance

Question 38

What is another word for coefficient of determination? (explained variance)

Answer 38

R squared

Question 39

When do you use R-squared?

Answer 39

When you are evaluating a single model

Question 40

When do you use R-squared adjusted?

Answer 40

When you comparing different models

Question 41

What happens to R-squared when you keep adding more independent variables?

Answer 41

R-squared go up

Question 42

What does parsimony mean?

Answer 42

a simple regression model is referred over a complicated regression model. 2 or 3 independent variables is better than 10.

Question 43

What is another name for the significance?

Answer 43

P value

Question 44

What is the classical cutoff value for significance?

Answer 44

below 0.05. it means that the statistics is significant. = r squared is significantly different from zero = good regression model. above 0.05 = the factor does not say anything

Question 45

What can we say about r-squared when the significance is below 0.05?

Answer 45

r squared is significantly different from zero. its a good regression model