0
$\begingroup$

I'm taking an introduction to statistics course and in attempt to reinforce what I have learned I have fit normal distributions to oil open,high,low,close prices over past 10 years where the number of bins is 25:

enter image description here

What can I infer from these distributions ?

As the curve is relatively flat there is much variability in the prices. The data prices are not skewed and is approximately symmetric the mean and median prices are approximately equal. The data is also multi modal as there are multiple prominent peaks.

There is much space between each of the histograms, increasing the number of bins from 25 to 1000 renders :

enter image description here

As you can see this changes the shape of my distributions. Does this mean my previous assertions are incorrect ?

$\endgroup$
  • $\begingroup$ What exactly are you trying to estimate? Fitting normal distribution to this data clearly does not make sense since it is multimodal. You could try a mixture of normals. $\endgroup$ – Tim Sep 13 '17 at 12:46
  • $\begingroup$ @Tim I'm attempting to understand the data set to gain insights into what exploratory algorithms I can execute in oder to predict future price. Again this is just a learning exercise. $\endgroup$ – blue-sky Sep 13 '17 at 15:09
  • 3
    $\begingroup$ But this approach would lead you nowhere since there is no reason whatsoever to assume normal distribution for this data. $\endgroup$ – Tim Sep 13 '17 at 16:28
0
$\begingroup$

You cannot fit the normal distribution to oil prices, or for that matter any prices. The normal distribution is sometimes applied to the returns on prices. The reason is that prices are not stationary generally due to inflation and growth. You need the stationary series (at least) to apply the probability distributions, such as normal.

You cannot predict future oil prices without time series analysis. So, wait until you get to this topic. Even then you won't be able to predict them, but at least you'll see why.

Otherwise, there is a reason why in the textbooks the example are chosen the way they are. It may frustrate you that the examples are useless, and you'll want to apply the classroom knowledge to real life, practical examples. However, it usually doesn't work while you're on introductory levels of statistics, because the world is very complicated and asks for sufficiently complicated analysis

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.