Let's suppose I have weekly data available for making a forecast using a linear regression model. Let's also assume that the weekly forecasts can be aggregated to monthly ones. Would it make sense to compute the forecasts using weekly data and sum it up to monthly level, or would it be better to just aggregate the data to monthly level and then do the regression? Does it matter? The model done using the weekly data has a lower accuracy than the one that uses monthly data, but how would aggregating affect the accuracy?