I noticed that on the raw data set for tweets with specifics hashtags that i collected, whenever the date of the tweet occurred on December 31st, R considered it as the 53rd week of the year, however, should only have 52 weeks.
When I join another database (Google), which is correct, with the database extracted from twitter, R put these unmatched rows at the end of the dataset - which "broke" the timeseries chronological order. (you can see here) (https://i.stack.imgur.com/TxsTd.png) `
This mistake was carried out throughout the data analysis to a point that when I created a new column to add a date variable for my time series graph, these last rows that were out of order were assigned the wrong month - they were wrongly considered as month 1, instead of month 12.
Given such a mistake, in a further chunk, when I joined my main database (twitter) with the database of the firm's social media activity, the last two rows received a zero on the month column, which led to the error in the seasons dummy.
I need to adjust the chronological order to get rid of these miscreated fake leap years