I am not a data scientist, I work in different parts of science. Of course I understand that every field has its own jargon and internal expressions. But I keep wondering.
So, was Laurence using specific data science terminology?
For example:
What I would call gradient of regression line, he calls ‘trend’
What he calls ‘seasonality’, I would call periodicity (or periodic signal, etc.)
White noise is white noise, we agree on that
What he calls ‘autocorrelation’ I would call ‘ringing’ or ‘multipath’ or ISI (inter-symbol interference)
What I would call ‘fitted curve’ he calls ‘forecast’
What he calls ‘imputation’, I would call interpolation (to fill gaps) or extrapolation (to do the ‘forecast’)
I am not a data scientist too
However, I do agree with what you’ve thought except autocorrelation.
What I actually understand this as- Autocorrelation : Indicates the same variable changes over time t (may watch this) rather than different variables.
In ‘ringing’, a time-domain reflectometer (TDR) would measure the return time (2*k in the video) of the signal y. For practical benefits, this measurement signal is a narrow sharp impulse (a ‘click’ or a ‘delta function’) generated by the TDR’s trasmitter. On the receive side, we will see the original pulse and the delayed, diminished pulse.
For the ‘multipath’, in radio communication, if the signal gets reflected from various surfaces, it will reach the receiver later than the direct path. On old analogue TVs, this couple of microsecond delays are seen as ‘ghost picture’.
…and for ‘ISI’, is effectively the same, but for digital signals: and ‘echo’ of the previous symbol or previous data burst overlaps with the current one. There are various ways of mitigating it, such as this one.
Anyway, thanks for showing the video, it was very nice and clear!
