Time series: daily data and daily data without weekends I've got a daily data set on Bitcoin's price. Trading continues in the weekends, hence there is price data in the weekends as well. 
On the other hand, i've got another daily data set. This set does not have data for the weekends. 
I want to make a time series regression with the two sets but what should i do with the weekends? Should I replace the empty weekends with the average? Then, missing values will just be replaced by the average. 
Edit:
Im going to regress dataset X on the price of BTC. Thus BTC is the dependent variable.
Dataset X: no weekend data, workweeks=daily data
BTC: daily data on the price, also weekends. 
 A: Assuming you produce two models, one for predicting Bitcoin (BTC), one for predicting the other while always using both as inputs:


*

*For predicting BTC (imputing input only): Fill the last value from the work-week (e.g. Friday) as value for the weekend.

*For predicting the other (imputing input and output): You can do the same as above (will introduce a slight bias to predict values that appear on Fridays). Or you can consider discarding weekend BTC data and just treat the problem as time series without weekend.

A: Sorry to ask a question but I don't have enough rep to comment. Do you have to use this dataset? 
https://www.investing.com/crypto/bitcoin/btc-usd-historical-data
This dataset contains all close prices including weekends (you can adjust the amount of time but it does only have data going back to 2012 if that's a concern).
If not you should use imputation, in R there are many packages to do this, careful which one you choose though because your data is almost certainly autocorrelated. While I have not used it myself, the 'imputets' package is supposedly a strong choice for imputing the values in R, also consider 'amelia' (uses bootstraping, which would be better and less biased).  You could take a moving average to impute the data yourself if you prefer, but I think if you can use a dataset that has the actual values you should, what if volatility spikes on weekends, or conversely, there's less vol on the weekends because major firms that trade in the currency may not be actively trading? If you can't find a full dataset, a time series specific imputation would be most efficient and least biased, compared to an average. 
