I'm not an addict
Spam no!
Feedback yes!
Not a whole lot we can deduce, is there? Well, I suppose we can make the claim that I talked more in the first half of the time period than the second half, but that's about it. Well let's see if we can identify trends using the discrete time Fourier transform (DTFT). The absolute value of the DTFT is plotted:
Now this provides a wealth of information. For one, it's symmetrical around the halfway point since the data are [ obviously ] real-valued. There is a huge low frequency component in this data. Zooming in on the first 25 values:
The DC component is about 3.5 times higher than the next element in the DTFT, which is over twice as high as the rest. This tells us that the data are [ clearly ] all positive, and that there is a strong trend between the halves of the time period, all things we knew before. There are also some minor peaks around ~5 and ~10, which represent yearly and half-yearly trends, respectively. However, a larger peak is seen around 22, which represents a period of 1946 days / 22, which is pretty darn close to 3 months, indicating a seasonal trend in IM usage.
The rest of the DTFT data reveal a rather large peak approximately between 275 and 285. This proves that my IM habits depended strongly on the day of the week. Actually, if you look closely, you can also see the peaks for 3 month, 2 month, 1 month, and 2 week trends.
For completeness, this is the real part of the DTFT [ red ] and the imaginary part [ blue ] plotted together, without the first two elements:
