Correcting for Nonignorable Nonresponse Bias in Ordinal Observational Survey Data