The seminal paper Robert Litterman, José Scheinkman (1991) Common Factors Affecting Bond Returns. Journal of Fixed Income, 1, 54-61. https://doi.org/10.3905/jfi.1991.692347 used the history of daily changes in observable par rates. Some people prefer to bootstrap & convert to zero/spot rates. What are the advantages/disadvantages of each approach, abd are there are any choices that one should consider?

What are pros and cons of using par rates v. zero rates for principal component analysis of daily history of a yield curve?
Dimitri Vulis
