Many times traders used to tell their process of setting up the trade as scientific and blah blah blah. But to call something as scientific, every technique has to go through a proper scientific research process. It also has to stand its ground to rigorous review by a tough human being and also the process is subjected to tough acid bath of statistics. The scientific research process unlike an algorithm of computers is a linear one. Here is steps to follow in scientific research process.
Steps in Scientific Research Process:
- Problem Definition
- Review of previous works related to the problem.
- Setting up of research apparatus and Formulation of Hypothesis..
- Running the research and collecting the data.
- Analysis of Data collected and validation of hypothesis.
- Presentation of the result and expert review.
The above list is how a research happens. These steps can also be seen in different mythological context too. The 6 steps corresponds to What How and Why’s of a thing (Also read: what how and why framework here.).
Step 1: Problem Definition:
The first step in research is definition of the problem. The problem is nothing but an telling of whats really happening in the environment. Problem definition can be like “Why sales of sunfeast biscuits is so low in city of Dharwad?”. The problem can also be “how to measure oversoldness of a stock” etc. In short whatever the so call pundits call as scientific can be considered as problem definition. The key criteria of good problem definition is the problem should entirely based on subject and should not have any reference to statistics.
Step 2: Review of Previous Works:
The review of previous works sounds more odd as many feel that their works is unique and will not be covered by previous researches. When a review is done it sheds light on variables and the interplay between them which can also have bearings on your research. Also the review sheds light on biases that may creep in. If you haven’t read on efficient market hypothesis and started off with your research on trading techniques, its expected to be one sure fire biased work. Its the review works which can differentiate a good researcher from a bad one. Review also helps in understanding the subject properly, and help in forming hypothesis (its obtained by subdividing the problem statement into small measurable chunks).
Step 3: Setting Up and Hypothesis formulation:
In this phase we set up the research process apparatus i.e. deciding on things to measure, how to measure etc…
First step of it – the problem statement is subdivided into chunks called hypothesis. The hypothesis is a set of 2 statements which describe small set of problem . One of the hypothesis is called as Null Hypothesis and other one is called Alternate Hypothesis. Null hypothesis states that there is no relation between the variables, where as alternate states there is relation between the variables. And every hypothesis is made of variables which are measured to know the validity of the which of the hypothesis holds true. The main feature of hypothesis is that only one of them can be true.
Apart from preparing hypothesis you are also supposed to identify the proper audience also called as sample. (sample is the set of people on whom the research is done, Sample is selected from population hence its subset of population.) The process of selecting sample is called sampling and is decided in this step itself.
You must also decide how go about collecting data without letting bias to creep in. You are segment sample to allow pure randomness, so that biases due to concentration doesn’t creep in while collecting data. Bias can also question the validity of research hence one needs to be careful about it. You can call this step of sampling, and prep work as preparation of statistical model, as it lays out model for data collection. Only after proper model is setup one can go about data collection.
(Know more on statistical hypothesis testing on wikipedia, it explains about model too)
Step 4: Data Collection:
Data collection is the smallest step of the research process. Here the research is conducted and data gathered. In case of research related humanities subject often times the questionnaire is submitted to people to answer. There are other types of experiments to collect data like focus groups, blind tests etc. for human participants. For non humanities subjects like physics etc, the research is ran on machines and data is captured. The machine on which experiment is conducted is part of experimental set up. The data collected from it is called as sample. The data analysis step follows the data collection.
Step 5: Data Analysis:
This is the step where acid bath of statistics happens. This is one of longest and important step of scientific research process. Here the statistical analysis of codified data happens. Based on the statistical analysis the results are published.
Once the data is collated from questionnaire, it needs to cleaned and be made machine ready. For example if questionnaire had rating scales then the answer would be like strongly agree to strongly disagree. These kind of likert scale answers cannot be fed into machines directly hence they need to codified like strongly agree = 5 and strongly disagree = 1. (One has to note that the coding strategy is pre-decided in research set up phase, here only its implemented based on it.)
After the coding and data entry is over Software like Matlab, SPSS, SAS are are used by researchers to run the various statistical analysis on the data. Things regression testing, factor analysis are done in this phase to validate hypothesis. The analysis is called as hypothesis testing, since these analyses are done to validate which hypothesis is true. Based the result spewed out by the software which hypothesis is valid is determined.
Step 6: Results & Review:
The final step is of publishing the results and its subsequent review by an expert. Based on the experiment conducted the valid hypothesis is collated and result is published. If one can recall, the data collection of gravity waves ended way back in September of 2015 still it took a lot of time to run the analysis and publish the results. Once the results are published it has to go through a panel which vets whether the research was done in unbiased way, once its fairly confident of absence of biases then the research is published.
There are some caveats in this scientific research process. In the above mentioned scientific research process, the researcher does the experiment to confirm his gut feeling. For example the discovery of gravitational waves was to confirm whether the gravitational waves exist or not.This kind of confirmation of gut feeling is called as confirmation research. There also another branch of research called as exploratory research, it follows a different scientific research process but instead of hypothesis testing it just measures the variables and tries to build relation ship between the variables. This research falls under the realm of big data.
The above said steps describe the scientific research process. All the statisticians in a company follow the above mentioned steps when they do their market research or other kinds of R&D works in a company. To know more the above mentioned scientific research process in depth you can read a book by Thomas Davenport called as “Keeping Up With Quants”.
Here is amazon links to its hardcover and kindle editions.
If you are trader then remember that trading is art, its not science. Its so because the first rule of trading is “Be Flexible” and science is never flexible. Since the above steps are very laborious, dont call every one of your gut feeling as scientific, you can call your project in MBA as scientific though. ;).