Short steps on how to read a paper. Part 6: In search of the lost hypothesis.
This next post in our series is on the importance of the hypothesis. Both readers and investigators often overlook this. Hence the title of this post is called “in search of the lost hypothesis”.
When you read a paper, you should look for the hypothesis just after the aims and objectives.
In short, the hypothesis should be an unambiguous statement that describes a relationship between two or more the variables that the authors are testing in the paper.
You should look to see if it has the following qualities:
- Clear and specific
- Relevant to the research question
- Be testable. This concept is crucial to your reading of the paper.
In clinical research, authors often state the null hypothesis. The reason for this is statistical tests enable us to accept or reject the null hypothesis. It is helpful to look at some good and poor examples of hypotheses. I want to go back to the aims of a study that I mentioned in my last post. This study was looking at “magic brackets”.
An inadequate hypothesis would be
“Magic brackets are not better than conventional brackets”.
This statement does not satisfy the criteria above.
A better one would be:
“There is no difference in treatment duration in months between magic brackets and conventional brackets when treating 11-16-year-old children”.
As with the aims, authors need to be clear, and this information is essential for the interpretation of their paper. If it is not clear, there is a danger that the report will drift and we will achieve nothing by reading it.
Next week we will look at the dreaded statistics! Padhraig is going to write this one…
Dear Kevin,
You’ve hit a wonderful series with your reading papers theme. I give our residents a course on reedin’ rytin’ but not rithmeticin’
And they even review manuscripts for the Angle Orthodontist; occasionally AJO.
You are too wonderful.
Hope you will have a memorably wonderful Christmas.
Dear Kevin,
This pertains to Karl Popper’s concept of falsifiability in his book on objective knowledge. If it cannot be tested or measured it is difficult to regard such theories as sound and robust one that has been tested. Popper also preferred bold theories and challenges rather than incremental tests that added little to knowledge.
There can be tension between a bold but difficult to test theory versus an easy to test model.
Your examples refer to NHST. Movement away from most aspects NHST should be encouraged as such papers are prone to be misinterpreted and misleading. We should be very cautious when reading articles which claim statistical significance, especially when the protocols and M&Ms have not been published prior to the experiment.
Our speciality’s science has been served very poorly by poor statistics and poorer interpretation, often through ignorance.
Dear Prof Kevin O’Brien,
One could consider using PICO to formulate the hypothesis.
P – Population
I – Intervention
C – Control
O – Outcome
Using the magic bracket example in the above article:
There is no difference in treatment duration in months (outcome) between magic brackets (Intervention) and conventional brackets (control) when treating 11-16-year-old children (population)
Your hypothesis has thus fulfilled the PICO criteria.