1

I have a dataset showing the yearly stock returns for two investment styles (growth and value investing) as well as a column showing the difference between the returns for each year. I have attached a picture below.

I want to do a test to see if the difference is significant for each individual year, such that the test fills out the area highlighted in blue.

Which test do I use? And if I use that test, what is the Excel command I need?

Spreadsheet

mkt
  • 11,770
  • 9
  • 51
  • 125
dakofsta
  • 29
  • 4

1 Answers1

3

This seems misguided to me, but perhaps more importantly you cannot test if the two strategies are different in each individual year with this data.

To simplify this a bit: you only have one data point for each strategy in each year, but testing for a difference requires a set of data points so that you can compare the averages (usually the mean). If you had data for each of the strategies from each year from many different countries, for example, you would be able to test whether the strategies are different. I would still not recommend doing so, however! There are almost certainly better ways to address your underlying question.

I would strongly recommend reading more about hypothesis testing, statistical significance and p-values before proceeding. There are a number of relevant threads on this site, including:

Why is "statistically significant" not enough?

What is the meaning of p values and t values in statistical tests?

When to use Fisher and Neyman-Pearson framework?

mkt
  • 11,770
  • 9
  • 51
  • 125
  • 1
    Thank you very much! – dakofsta Apr 24 '20 at 18:37
  • @dakofsta Sure thing! – mkt Apr 24 '20 at 18:38
  • Since I'm using this for my dissertation, I would like to use some form of statistical test on the whole dataset. I could do a one tailed T test to see if the averages for both means are significantly different, but are there any other tests you would recommend? I know it is cheesy but I get additional marks for using statistical tests in the data – dakofsta Apr 24 '20 at 18:38
  • @dakofsta You shouldn't do that because data points in time series are not independent, which is an assumption in t-tests (and many other simple hypothesis tests). I recommend searching this site, because this has almost certainly been addressed before here. But if you cannot find an answer, post this as a new question with a plot of your two time series. – mkt Apr 24 '20 at 18:42
  • I've tried to find the answer all day and haven't found any useful tests :( – dakofsta Apr 24 '20 at 18:49
  • @dakofsta Post a new question, as I said. – mkt Apr 24 '20 at 18:53