Symbolbild: Texte aus einem Wörterbuch ausschnittsweise auch auf eine CD gedruckt

Bachelor's thesis

Symbolbild: Texte aus einem Wörterbuch ausschnittsweise auch auf eine CD gedruckt
Image: Jan-Peter Kasper (University of Jena)
This page has been machine translated.

Bachelor propaedeutic course

Information on the organisation of the bachelor's preparatory course at the Professorship of Psychological Methodology can be found in this document: Pre-semester preparatory course for the bachelor's dissertationpdf, 178 kb · de.

Bachelor thesis topics

The following Bachelor's thesis topics are currently offered by the Professorship of Psychological Methodology (further or own topics on request):

Topic Topic description Are you interested? Please contact: Further supervisors
Non-orthogonal ANOVA with repeated measures Analyses of variance (ANOVA) with repeated measures are often used to test the significance of mean differences between different groups over time and to make statements about main or interaction effects.
make statements about main or interaction effects.
There are four different methods for carrying out this examination (type I - IV). The correct selection is particularly important in non-orthogonal research designs in which, for example, there was no randomised group allocation and individual cells are unequally populated. The users of the analysis of variance are often not even aware of these sometimes serious differences in the methods and they follow the respective (different) default settings of the common statistical programmes in their analyses.
In a bachelor/master thesis, the effects of an ill-considered ANOVA selection on the effect estimates are to be analysed with the help of a simulation study. An illustration, e.g. with the help of a Shiny app, would be an optional addition to the thesis. This could build on existing preparatory work|preparatory activity on this topic.		 Linda Gräfe de  
Rules of thumb for the use of non-orthogonal analysis of variance Analyses of variance (ANOVA) are often used to test the significance of mean differences between different groups and to make statements about main or interaction effects. There are four different methods for carrying out this examination (type I - IV). The correct selection plays a role particularly in non-orthogonal research designs in which, for example, there was no randomised group allocation and individual cells are unequally populated. The users of the analysis of variance are often not even aware of these sometimes serious differences in the methods and they follow the respective (different) default settings of the common statistical programmes in their analyses.

There are no rules of thumb as to when non-orthogonality actually becomes problematic.
A bachelor's dissertation will use a simulation study to investigate which factors have an influence on the ANOVA results,
when a non-orthogonal design is present.		 Linda Gräfe de  

Review of affect scales

Current affect influences one's own affective experience and behaviour in subsequent situations. Various response scales can be used to assess current affect. While initial differences between analogue (e.g. slider) and discrete scales (e.g. five response categories) have been reported, little is known about the differences in response behaviour between discrete response scales. A distinction can be made between unipolar response scales (e.g. affect "happy": strongly disagree - strongly agree), bipolar response scales with only the endpoints labelled (e.g. relaxed - tense, on a seven-point response scale without labelling the intermediate categories) and bipolar response scales with all response categories labelled (e.g. very unwell - unwell - somewhat unwell - neither - somewhat well - well - very well, on a seven-point response scale).

In the present work, a small survey is to be conducted self-employed (experience sampling method via a smartphone app), in which the same items are presented with different response scales. Differences in the responses to the items depending on the response format are analysed and discussed.

 Miriam Jähne de  

Heterogeneity in longitudinal data

The term heterogeneity refers to differences between persons (groups) that lead to different results for each person (group). Accordingly, the consideration of the heterogeneity of individuals has been gaining in importance for years. Last but not least, difficulties in replicating earlier research results using a comparable sample (e.g. the results are to be replicated on a new, comparable sample) or generalising earlier research results to a larger sample (e.g. the results are to be extended to a broader sample than the one previously used) can be attributed to the lack of consideration of heterogeneity between individuals.

Known approaches to take heterogeneity into account are the use of covariates or the use of classification methods (e.g. LCA, cluster analysis) that identify groups of people. This paper is intended as a literature review and aims to research and compare approaches for taking heterogeneity into account in longitudinal data. Limitations and potentials of the different analysis options will be considered.

 Miriam Jähne de  
Speciesism Speciesism is a construct from Ethics and describes an ideology according to which the lives and experiences of animals are considered inferior or less valuable than those of humans,
because they belong to a different species.
The question arises as to how this construct is treated in Psychology.
What instruments are there for recording it?
In what language are these instruments available?
What are the psychometric properties?
and much more.
As a bachelor's dissertation, a comprehensive literature review would be prepared to answer these and other questions.		 Lukas Knitter de  
Anomalies in dark triad bifactor models Eid et al. (2017) declare that the classical, orthogonal bifactor model may only be fitted to data that fulfil certain conditions,
that fulfil certain requirements.
Questionnaire data, which is common in Psychology, often does not fulfil these requirements.
If such data are nevertheless fitted with this model, so-called anomalies often occur, results that appeal|to objects either in the conceptualisation of the bifactor model (e.g. loadings <= 0) or in general mathematical principles (negative variances).
Nevertheless, the classical bifactor model has become increasingly popular in recent years.
Among other things, the personality traits of the Dark Triad are often fitted with this model.
As part of the bachelor's dissertation, a systematic literature review will be carried out to find out how often anomalies occur when data from the established Dark Triad instruments (Short Dark Triad, Dirty Dozen) are fitted with a classic orthogonal bifactor model,
orthogonal bifactor model.		 Lukas Knitter de  
Systematic literature review on identification problems with extended bifactor models Confirmatory factor models with general and specific factors
(e.g. bifactor models, hierarchical factor models, or G-factor models)
are very popular in psychological research and practice.
In particular, bifactor models are used to analyse multidimensional data
in Psychology (e.g. in intelligence research).
A key objective in the application of bifactor models is to analyse the
general and specific factors as predictors (independent variables) for the prediction of external
prediction of external criterion variables (dependent variables).
However, this often leads to identification problems (i.e. non-estimability of the model and model parameters).
of the model and the model parameters). In the bachelor thesis, a systematic literature
literature research on this topic.
This should also include a search for relevant solutions,
which may be proposed in the literature.		 Tobias Koch de Maximilian Bee de
Literature review of model quality criteria in Bayesian statistics In this BA, a systematic literature review will be conducted to identify how model performance critiques and comparisons in Bayesian analyses are reported and evaluated in psychological publications. While there are established cut-off criteria for frequentist model comparisons, model comparisons for Bayesian analyses are not yet widely used in psychological research. The literature review focuses on model comparison criteria such as Bayesian Information Criterion (BIC), Watanabe-Akaike Information Criterion (WAIC), Leave-One-Out Cross-Validation (LOO-CV), and Bayes factors. Fabian Münch de  
Simulation study on measurement invariance in confirmatory factor analyses with frequentist vs Bayesian model tests In this BA / MA, a simulation study will investigate whether (and if so, why) there are different results when testing measurement invariance in structural equation models using (1) frequentist test procedures such as likelihood ratio tests and (2) Bayesian test procedures such as leave-one-out cross-validation or Bayes factors. The analysis effort can be adapted depending on the BA / MA, but good knowledge of R is required. Fabian Münch de  
Data analysis on a self-selected topic with data from the family panel (pairfam) or socio-economic panel (SOEP) In this BA / MA, various questions on data from two large German panel studies can be analysed. Many personality traits and other psychological constructs are recorded in one or both studies (e.g. Big 5, loneliness, attachment styles, ...). The question should be investigated with a suitable statistical model (HLM, SEM, ...), with the focus on modelling. Fabian Münch de  
Influence of delays in the experimental setup on reaction time experiments The field of human-computer interaction (HCI) is interdisciplinary and combines in particular
combines Computer Science and psychological research in particular. As part of
Central Office for Teaching Evaluation (ULe) we would like to offer the best possible surveys.
possible. This includes making (online) experiments as reproducible and valid as possible.
measurements as possible.

This bachelor's thesis topic is intended to build on research work already begun on the
influence of unintentional delays in reaction time studies. A first
experiment has shown that reaction time studies are relatively robust against unintentional
delays. However, the experiment showed potential for improvement in the experimental setup.
Within the bachelor's dissertation, the experiment is to be improved in this respect
and conducted with test subjects. The statistical analysis of the collected data
is part of the bachelor's dissertation. Basically, the bachelor's dissertation is carried out in a small team.
team.		 Thomas Prinz de Linda Gräfe de
User-friendliness of different rating representations The subject area of human-computer interaction (HCI) is interdisciplinary and
combines Computer Science and psychological research. Within the framework
Central Office for Teaching Evaluation (ULe), we would like to offer the best possible online questionnaires
for a wide variety of surveys. This includes creating representations of
items (questions) to the test.

In this bachelor thesis topic, different representations of rating items are to be tested
for their user-friendliness, especially on small end devices.
devices. For this purpose, it is necessary to find scientific questionnaires on the topic of usability
in the literature and to check their suitability. With the help of a
questionnaire is then selected, a survey is conducted. Fundamentally
be investigated as to whether rating items should be presented in a horizontal or vertical
vertical arrangement. Furthermore, for a vertical arrangement
different representations of the "buttons" are possible: from circular to buttons.
The statistical analysis of the collected data is part of the bachelor's dissertation.		 Thomas Prinz de Linda Gräfe de
Course evaluation (LVE): Differences between paper-based vs. online surveys At FSU Jena, around 800 course|classes are evaluated each semester with around 15,000 questionnaires. Over the last 15 years, an extensive data set has been collected. In addition to the students' assessments, the lecturers' self-assessments are also available. In addition, there are some potential covariates for students (sex, semesters enrolled in, study effort) and lecturers (sex|gender, age, specific degree programme/institute, academic degree, teaching experience, type of course (lecture, seminar, tutorial), course size).

The course evaluation is a complex multilevel design, which entails some special features in the data structure:
(a) The LVE is an external assessment of several students on a course|class (multiple rater).
(b) Students give an assessment of a course|class and are simultaneously nested in a higher unit (a so-called cluster), namely a course|class (multilevel setting).
(c) Lecturers and students make multiple contributions to the data set: Students provide an evaluation for several course|classes. Similarly, lecturers have several course|classes evaluated (cross-classified multiple membership). This violates the assumption of the independence of observations (iid assumption, independent and identically distributed).
(d) The student assessments with the PELVE instrument are carried out on five-point rating scales. The distribution of the variables is skewed and therefore cannot be assumed to be normally distributed, interval-scaled variables.

These particularities must be taken into account when selecting the analysis methods. The data should be prepared and analysed in R (in particular with the lavaan package).

In addition to comparing the survey format (online vs. paper), further questions are possible:
* How does the students' assessment differ from the lecturers' self-assessment?
* Are courses with several lecturers (co-teaching) better than courses with one lecturer?
* Latent class analysis to classify students: Are there latent classes of satisfaction?
* How do lecturers develop over time?
* How does the evaluation behaviour of the same student change over time?
* Influence of covariates (sex|gender of student, sex|gender of lecturer, age of lecturer, semesters enrolled in, degree programme, course size, type of course, etc.) on the evaluation results

Own questions are welcome.		 Anja Vetterlein de Jan Plötner de

 

The following bachelor thesis topics are currently being worked on:

Author Topic Topic Description Supervisor Other supervisors
  Simulation study on regularisation methods in structural equation models   Christian Bloszies de