S93B / Portfolio

Personal portfolio for learning Python
0 stars 0 forks source link

Creating of political trust scale #5

Closed S93B closed 3 weeks ago

S93B commented 1 month ago

The idea here is to combine the multiple indicators of political/institutional trust. Depending on the outcomes I might choose to combine the more typical political trust indicators into one measurement, or I might include trust in e.g., government and legal system for a more institutional trust measurement.

  1. Using Principal component analysis for reduction of trust dimensions. It has been awhile since I have done this test. Besides figuring out how to run it in Python, I also need to dive into the theory somewhat. Luckily I have lots of books from my study.

  2. Cronbach alfa for reliability testing

S93B commented 1 month ago

Factor analysis extracted two factors, the items from the first wil be used for the political trust scale.

Cronbach alpha = 0.92 indicating high internal consistency.

Problem: Sample contains about 121 rows that have at least one NaN. Decision: Decide whether to remove the rows, or perhaps calculate answers based on similar respondents. For the latter I will need to figure out how viable that is. I have a vague memory that I have done this before, cannot exactly remember.

My opinion is that it does not hurt the data to remove the rows, however it seems like fun to think of a different way to tackle this.

S93B commented 4 weeks ago

I used k-nearest neighbors to fill in the missing data. Combined the political trust variables into a mean scale.

sidenote: I ran the factor analysis on data without the missing rows. I don't expect much difference in results, but still important to note