Daily data, different dates I have downloaded two different daily data sets, both for the same period. Together I want to use them to construct data for another variable. However, the problem is that the dates do not exactly correspond. What would be the best approach to combine the data without having to convert it to monthly data?
Edit: I noticed that the question is slightly ambiguous in its current form. What I want as an end result is two series (not one) with the same dates (I will later use those to create a third variable). Each time a date is present in one of the two series but not in the other I want for instance an interpolated value.
 A: In MATLAB the easiest way is to use financial time series objects. You create fints for each series, then use merge function. It has several interesting options, such as to either use 'intersection' or 'union' on dates, and you seem to want the latter. The default "union" option will include the union of all dates from both series. When the data is missing for one series it'll be NaN. You'll have to deal with NaNs afterwards. For instance, if you call mean on a series or cov on a pair of series, then you'll get NaN as answers. In this case you could use nanmean and nancov. Similarly, your regression or other models may fail encountering NaNs. Some methods such as state-space models or ssm will work with NaNs just fine.
The "intersection" option of MATLAB's merge will include only dates present in both series, so there are no missing data, but at the cost of omitting partial observations. 
In case it wasn't clear: the merged fints object will contain two series inside. Their names can be obtained by calling ftsinfo function. There's a host different methods to retrieve data from the object, such as fts2mat, which will extract the series into a matrix, where each column is a series and rows are observations etc.
Additionally, there are functions such as tomonthly, toweekly' andtodailyto downsample to monthly if you choose so. For instance,tomonthly` would down sample your series to monthly, setting the date at the end of the month. By default it accounts for the business days vs. holidays and weekends, but you can change that.
@Achim also reminded me of the interpolation function called fillts. It interpolates missing observations using different methods such us linear. I would be careful with any sort of imputation though to make sure that you are not silently artificially propping up your sample size.
