I am conducting a research where i need to calculate z-scores and some other market ratio. I selected about 316 manufacturing firms listed at KSE for 15 years range from 1998 to 2013. I collected daily stock prices and took mean for a particular year observations. i also need to calculate average daily returns and stock volatility. However, there are number of companies having no activity or very few days activity (like 10 or 15 days) for a particular whole year. Including such year observation can provide biased results. How can i deal with this problem?
please suggest and recommend actions with references.
When you do backtesting, try removing such companies. Check Ernie chan's also trading title. I think more is given in Chapter 4
You can deal with this issue using SAS. First you calculate 'mean procedure' then apply 'if' statement with minimum observation. If you have any query then feel free to contact me.
You have large data considering the years understudy. Also you have high number of companies which will give you much headache than you can imagine. You must have set parameters for your data collection from the beginning of your research. Between the years under study, there are newly established companies and as well there are some that have bankrupt or merger and acquisition have occurred. All these must be your initial assumptions.
However, there is still a room for improvement. You need to reverse or reset your initial assumptions, set qualifying parameters that each firm must meet before being included in the measurement, eliminate firms that life spans fall short of the years under study i.e. 1998-2013, any merger and acquisition between the year 1998-2013 should be given special conditions either before or after the merger etc.
The results of the above idea would reduce the number of firms under study. Eliminate firms that have lower activities etc. Hopefully, this would help you.......
Thnx for all responses.
My problem is that what cut off point (i.e atleast how many number of days activity) i should use. plz i will appriciate your answer with references.
My answer is yet like my last opinions. Using daily data for 15 years is too large and you will end up facing lots of problems. The data would be too large and I don't know the type of software you will use to run the data.
It is better you use monthly or quarterly data. This would lessen your burden considering the number of firms under study. Please find references for this answer because this is a research experience which individuals acquired over long period. Even there are other assumptions that are with references.....
Hi, I believe it’s better to exclude the specific companies. That helps you to achieve normality to your sample. After that you have also to restate your sample and period of examination. It is more suitable to set different time frames (or time classes) in case that you still want to examine the initial number of companies because different economic factors may affect your sample significantly. Finally, E-views could be suitable for your analysis.
you can probably use a minimum no of observations in the sample based on the statistical tests you envisage. Again for your subject matter I prefer E Chan or you can refer asset prices etc by Stephen Taylor or just apply CLT
Hi, removing some companies base on related economic parameters on a seasonal time period is suggested by me, and E-views or Spss for the analysis...
- Badges
- Science topic
- Similar topics
- Books