Crunching Numbers-They Can Be Crisp!

Looking at your data- crunching numbers while retaining crispness

You have done your study…the hard work is done.

You conceived an idea, wrote up a protocol, submitted it to an Institutional Review Board or an Ethics Committee, organized logistics for the study, recruited subjects, collected information, filled up case record forms.

You have all of the information in an electronic format (makes sense in these times when everything is electronic- we await reports of digital babies) and are now ready to do the next step.

Find the answer to your research question from the data you have collected.

There are several ways to do this

  1. You try to find the answers yourself
  2. You have an expert in statistics help you find the answer
  3. You wonder why you decided to do a research study.

Option 3 is not necessarily helpful at this stage (other than to try and recreate a feeling of genius) and hence we shall not discuss that further.

Options 1 and 2 relate to statistics.

 Statistics.

You either love it or hate it.

You either keep chattering about p values, tests, probabilities, odds, formulas, equations as if you are describing your long lost love…passion, respect, reverence and an illicit pleasure all evident in your voice….

OR

 You become transfixed as if you are on the top of Mt Everest without oxygen and with only a Yeti for company…a Yeti that wants you to pick fleas from its body!

Is there a middle path?

Why do we need statistics?

We need statistics because it is impossible to collect information from an entire population or from all persons of interest.  We collect data from certain subsets and hope that we can extrapolate that information to a larger population. We sample the population to collect information and derive estimates from that sample.

Statistics helps us derive estimates from a sample and to determine if the answers we obtain are indeed “true”. Truth is relative, and hence the answers we obtain are relative and usually have a range (of values, of conditions, etc) within which they remain the truth. Sometimes the values may lie outside the range of perceived truth…outliers. We will look at outliers later.

Ok, I need statistics to look at my data….

 What is the first step when you start looking at your data?

 The first step is usually the toughest step…..take a deep breath and LOOK AT YOUR DATA.

Do not jump into your software of choice and start crunching numbers. Remember, headless chickens run well but not for long.

Look at your data.

  1. Check each variable- its name, its structure.
  2. Recheck its relevance to the question
  3. Are there missing data?
  4. Do we have to clean the data? (example, a number entered as a text)
  5. Consider the best way to answer the research question

The ease of your analysis is proportionate to the time you spend looking at your data.

The time you spend looking at your data depends on the time you spent planning your data collection. If you planned well before starting the study, you do not have to spend much time looking at the data you collected.

Your analysis becomes much easier if you spend more time looking, understanding, cleaning your data prior to analysis.

Figure out the tests that can help you find the “truth”. Actually, we recommend you do this before you start the study. Have a Plan A, and then keep Plan B and C ready.

There are several statistical tests and several levels of complexity.

Start Simple. We do not really need complex models for many things. Complex Models, when not necessary, complicate matters. The more complex a model, the less likely it can be translated easily into practice.

How do we choose the tests?

  • Consider the assumptions that underpin each test. Each test can be used in certain situations- when certain assumptions are met.
  • Consider the distribution of the data and population from which data is collected- Normal distribution or not, variance, parametric or non parametric; random selection, purposive, representative, paired, unpaired design and more.
  • Consider the structure of the Variables- dichotomous, ordinal, continuous, categories etc
  • Identify the Independent and Dependent Variables
  • Consider the question you are trying to answer- FOCUS on that.
  • Consider Factors that may influence or impact on the question (example- maternal malnutrition and fetal growth)
  • Consider the answers you want to communicate.

 

There are several approaches to wet your feet in this.

We can do workshops, listen to interactive lectures and attend classes.

What we recommend is learning by doing, trying to answer the questions that interest you, mentoring you along the way. Consider it as 24X7 support to walk you through the common tests- why do we do those, how do we do those (for those  who would like to know), and how do we interpret those, what are the strengths and limitations of each test.

It is important to learn to read the results and understand the interpretations even if you do not do research.

Statistics are commonly reported, however, they are not always appropriately used or reported. When inappropriately used, they add meaningless noise to the results!

Keep watching the site for more information

Post written by Research Action Group- AMMA Education Research Foundation