# Drop data Vs fill data. Which one least hurts the integrity of the data?

I have a dilemma for an analysis I'm currently on. I doing some GARCH modelling of bitcoin and a fiat currency.

There are some null values with the fiat datasets in comparison with bitcoin data as cryptos trade on weekends. The R package I'm using rugarch won't accept null values

Now, I believe I have two choices:

1. Drop weekend rows of bitcoin
2. Fill weekend rows of the fiat currency

My question is, are there statistical justifications for either option and which option least hurts the integrity of my model?

There are plenty of imputation methods but the best is to drop data if you have enough data...

Imputation methods -> http://www.stat.columbia.edu/~gelman/arm/missing.pdf

Example of R package -> https://cran.r-project.org/web/packages/imputeR/imputeR.pdf

Another idea is to use algorithms that can deal with missing values such as XGBoost for example...

For your special case of time series, see -> How to fill in missing data in time series?

And the R package : ImputeTS -> https://cran.r-project.org/web/packages/imputeTS/vignettes/imputeTS-Time-Series-Missing-Value-Imputation-in-R.pdf

In a nutshell, if you can drop periodic data, you may adjust your interpretation either try imputation...

• The model works fine by dropping the weekend data. I am only worried if that invalidates the accuracy of my model and it's predictions. That's why I'm on the lookout for statistical justifications that the drops doesn't invalidate my results – Mysterio May 4 '18 at 9:46
• -1. I think your answer is missing a crucial point: dropping data changes the time dependence structure that is an essential feature of time series data. Moreover, your answer would better fit as a comment as the links and terms you provide do not answer the question directly. The direct answer to the actual question is the best is to drop data if you have enough data, but that is quite short and needs some elaboration and justification. @Mysterio – Richard Hardy May 4 '18 at 11:15
• @RichardHardy I know that that's why I have the dilemma. I'm looking at how to overcome this dilemma. What do you recommend? – Mysterio May 4 '18 at 11:18
• @RichardHardy any short answer or links that could help? – Mysterio May 4 '18 at 11:27
• @Mysterio, sorry, cannot remember any good sources right now. – Richard Hardy May 4 '18 at 11:38