The following graph shows the
Nasdaq composite, which consists of many technology
stocks.
Notice that the x axis is time, and the y-axis is the price.
Notice that the y-axis is unequally spread out.
Before 2000, you can see that even though there were many
fluctuations, the stock followed a clear upward trend.
But, it is hard to tell how well the data would fit the
trend line Excel would draw without the r2 value.
Excel can always find the best fit line, even in data that
doesn't seem to follow a trend (which would then be reflected
in a low r2 value). Looking at the graph from 2000 until it ends,
we see a steep downward trend.
The trend line over the entire time is still upward.
The following shows an Excel stock graph of monthly data
from 1998-2001. Just like your own stock graph,
the x-axis is the month (on your stock graph it is the day),
and the purple bar chart show the volume
(which is read by going to the numbers on the left).
The black and white boxes look like boxplots, but they are not.
They show the open, high, low and close prices
(high is the highest price a stock reaches in a given day,
and is at the top of the corresponding graph for that day) by going over
to the numbers on the right.
We can see from this graph that volume would not be a good predictor
of high. If volume was a good predictor of high, we should see
some kind of correlation between them:
In fact, we can see all sorts of contradictions to a reasonable correlation
between high and volume:
for example, from months 25 to 26, we see that the volume increases
as the high increases, but from
months 35 to 36, we see that the volume increases as the high
decreases (look carefully at the declining high, which can be seen
atop the purple bars).
There are many such similar contradictions to a consistent
trend in this graph.
Hence, we expect volume to be a poor predictor of high.
This intuition is reinforced in the Excel linear regression scatter plot
of Volume and High.
In this plot, volume is on the x-axis and high is on the y-axis.
Excel plots the best fit line, but this does not fit the data very
well because there is no clear trend.
As we can see, R2 is .1654 = 16.54%, which is a weak predictor,
according to p. 203 in HDYK.
You will use similar reasoning in your own stock graph in Monday's
lab.