M22 - Structural Equation Modeling

Question 1

SEM in a nutshell

= tries to reproduce an observed …/… …. of measured variables as … as possible by …. … … (and intercepts) in a … set of (usually …) … … derived from theory.

A good model:

Answer 1

SEM in a nutshell

= combines path models and confirmatory fachor analysis

=tries to reproduce an observed VARIANCE/ COVARIANCE MATRIX of measured variables as ACCURATELY as possible by ESTIMATING REGRESSION WEIGHTS (and intercepts) in a LIMITED set of (usually LINEAR) REGRESSION EQUATIONS derived from theory.

A good model:
Parsimonious
Theoretically justifiable
Reproduces the underlying corr matrix based on the constraints imposed

Question 2

What SEM can do

Answer 2

Its a combi of statistical data and qualitative causal assumptions

• graphical way to present an assumed theory
• estimates relship in the assumed theory
• estimates the causal effects of 1var on another (positive, negative, zero and signicabt)
• proves if the experimentally changed var actually causes an outcome

Question 3

Ocean model

Answer 3

The big five
Openness, conscientiousness, extraversion, agreeableness, neuroticism

Model that characterizises personality into these dimensions

By microtargeting, means measuring people’s personality from their digital footprints based on ocean model
–> manipulate election

Question 4

Importance in conjoint analyes

Answer 4

• the impirtance of an attribute is prop to the diff between. The part-worths of the best & the worst level of this attribute
• the impirtance values are normalized in such a way that their sum yields 100%
• the choice of levels of an attribute will in general affect the attribute’s importance

Question 5

What does it mean:
The importance of an attribute is proportional to the difference between the part-worths of the best and the worst level of this attribute

Answer 5

If i have 3 attributes
Each of them has differnet deltas between best and worst part-worths
The one with the biggest delta, has the highest importance

Question 6

Panel regression

If individual,

Answer 6

If individual, time-const effects are corr with the explanatory variable, then random effects regression is biased.
Fixed regression is still unbiased

Question 7

What SEM cannot di?

Answer 7

• prove if 1v actually causes anpther var
• prove the direction of causal order betw variables
• distinguish between models that result in identical corr patterns

Question 8

Regression model

Path model

Answer 8

Regression: several iv predict 1 dv

Path: several regr models are solved simultaneously

Question 9

Exploratory factor analysis

Confirmatory factor analysis

Answer 9

Exploratory: when a set of items is correlated, the items can be combined to yield a score

Confirmatory: test whether a set of items define a construct

Question 10

Regression models description

Answer 10

• only manifest var
• several iv predict 1 dv
• directed relationship –> factor loadings –> requires an error term

Question 11

Path models description

Answer 11

• combine several regr models
• only manifest var
• directed relship –> factor loadings–> require an errotr term
• indirect effect for x1 on y2 is mediated through: x1 on y1 and y1 on y2
• total effect: the sum of direct effect x1 on y2 and indirect effect

Question 12

Exploratory faczor analysis description

Answer 12

Factor loading represents whcih indicators correlate with the factor

No of factors such as to minimize factor loadings from indicators to one factor while minimizing cross-factor loadings

All factors may affect all items

Crinbachs aloha determines factor reliability

Question 13

Confirmatory factor analysis description

Answer 13

–> tests assumed structure

• priori defined no of factors and existence of corr
• only possible for reflexive indicators
• checks for unidimensionality
• typically an indicator loads only on one factor

Question 14

Advantage of path model plus confirmatory factor analysis?

Answer 14

Path: test relship between variables, but can only deal with observable variables

Cfa: tests cobstructs consisting of factors and. Indicators, and can use latent variables

Question 15

Reflective and formative operationalization of constructs

Answer 15

• reflective: the indicators reflect the construct (drunkenness)
• formative: the indicators form the construct (drunkenness)

Question 16

Conclusion of the study of OSS

Answer 16

• emergence of internet code abailable for free download from the internet is a nee opportunity of code reuse in software development
• intenet code is typically comes with licences demanding compliance with obligationd
• disregarding licences is a feom ob unethical behaviour
• intention to engage in the behaviour is driven by attitute towards it and perceived subjective norms
• an ethical work climate of conplying laws and codes has an indirect influence ob intetnntion being mediated by subjective norms
• attitude is influenced by perceived severity of time pressure and perceived punishment severity for both company and developer