Test to compare an effect on a variable on a daily basis

Question

I would like to know the effect of an external factor on the number of visits on a website during a period of 4 weeks (only search visits from Google). For this, I measured the number of visits every day through Google research. For example I have:

      Date Visits
01/01/2014    456
02/01/2014    423
03/01/2014    399
...

I did this for 4 months, including the 4 weeks when the external factor ran.

I would like to prove if there is any difference in the mean number of visits during these 4 weeks. My first idea is to run a t-test between two similar periods of 4 weeks, one will be the 4 weeks with the external factor, and the other will be a series of 4 weeks without. I have difficulty selecting the test: should it be a t-test paired comparing each day of the 4 weeks? I don't really know if I can consider my samples as independent as there is a weekly seasonality (on the weekends the visits are lower than on the week)

Answer 1

One way to work around this would be to remove the weekend data from both your samples of 4 weeks. Use a t-test or a non-parametric Wilcoxon if your data is skewed to check significance of the external factor. With only 20 data points though, the validity of your findings will be questionable.