En jämförelse av regressioner med binära utfall

• Main Author: Pettersson, Fredrik
• Format: Others
• Language: Swedish
• Published: Uppsala universitet, Statistiska institutionen 2020
• Subjects: Logit; Probit; Linear Probability Model; Monte Carlo; Probability Theory and Statistics; Sannolikhetsteori och statistik
• Online Access: http://urn.kb.se/resolve?urn=urn:nbn:se:uu:diva-451007

The purpose of this bachelor thesis was to compare three different methods for regression with binary outcomes. The three methods used for comparison are: Linear Probability Model, Logit and Probit. To compare the methods, data gathered from the World Value Survey when it last was done in Sweden in 2011 was used. The outcome variable in the creation of the models was whether the respondent preferred protecting the environment or economic growth. A Monte Carlo-simulation was also performed to strengthen the arguments in the comparison.  The results from the different models created was very small, but there are still differences. Two examples of the differences are the simplicity of interpretation between the models and errors that argues for not using Linear Probability Model under certain circumstances.