This week I was back to read the unfinished parts of The Truthful Art, Chapter 8 & 9 (Sorry for my random skipping…). The topic of chapter could be summarized as “how to use the time series plot efficiently and correctly”, ang Chap.9 focus on the scatter plot and its derivatives.
Time Series plot is a prevalent graphic, which is excellent at reveal the truth or the changes through the long-term period. When we analyze this type of plots, we should be concentrated on three things: Trends, Seasonality and Noise – Trends reflects where the variables go – up, down or just parallel with x-axis; Seasonality shows fluctuation in consistent or periodical condition; last, noise, I consider it as some irresistible factors which may influence the result. There is a conception called “moving average”, which I think is very useful when we need to compare two or three different but related variables (I suppose that it should be no more than three or the plot will be cramped.) in one plot and all of them are shifting.
Sometimes, percentages (e.g. rate of increase/decrease) are more important than raw value, like the percent of house selling at different prices showed in the instance of ‘House Marketing Analysis”, which exhibit a true change of house purchase. Here is a function:
Percentage Change = ((Each Score – Index Origin)/ Index Origin) *100
There is a logarithmic (LOG) plot, which fits things grows in explosion (or we may could say “exponential growth”). For example, the growth rate of bacteria, which reproduce in a low speed (compared with its later rate) at beginning but may explosion at the last few days. If we put all raw data in a graphic, the values of beginning days would be like drone of mosquito, but the end would be like sound of horn – nothing worth a digging. However, if we turn to analyze its rate of increase in a LOG plot, we could find more interesting things: not all bacteria engage reproduction every day, their lifetime could also affect the rate which could not reflect on raw-value based plot since the number is always increasing.
Time series could also be puzzling in several situations. First, if we draw a solid line when there are missing data, people will be misled as taking the value as the true counterpoint. So, we should the space blank or just put a dotted line. Second is ‘paradox’, like ‘Change in median wage”, how could the entirety still go up when its parts all go down? That could be explained for the different population at each educational degree. Third, the mix effort. According to the book, it is something like a Russian doll, the outmost size depends on its next layer – we may think the next layer is the last one, so we put them together as an association – however size of next layer also depends on its next layer. Finale is an introduction of a creative plot named as “horizon chart”, I like it definitely since it is shorter but still give the whole information.
I made content of Chap.9 in a simple flow chart.
Comments
Post a Comment