Dear Sir/Mam,
In my empirical study out of 5 independent variables only 1 independent variable is statistically significant, for remaining variables the p value is greater than 0.05. My doubt is whether can I proceed with writing the manuscript. Kindly suggest. Thank you
If you chose your independent variables from a theoretical model based on a hypothesis, then not finding an association is already interesting.
The first thing to look at is the effect size associated with each variable. This is because a p-value crunches two things together : the size of the effect and the precision with which it has been measured (effect size ÷ standard error –> p-value). So a large effect measured with low precision (large standard error) will have a non-significant p-value. If the effect size is large enough to be interesting but the standard error is large, then what you are looking at is an absence of evidence, not evidence of absence.
To study the variable properly would require higher statistical power, either from a larger sample size or from a measurement that contained less random error, or both. In a way, discussing the implications of a failure to find evidence is more nuanced and interesting than discussing a significant finding. It makes us ask if we are going about testing this association in the right way.
On the other hand, you may picked your predictor variables just because they were there, but there's no model or theory (some people hope that they can invent a theory based on whatever turns up in the data, which is a big mistake!). I'm assuming you wouldn't do something like this, but if you did, and one variable in five is significant, I'd just walk away. There's substantial chance that this is a false positive error.
Dear Balaji: the importance of the result of any variable is not necessarily to be statistically significant, but it is the true result whither it is significant or not. Suppose you have tested two variables such as temp. and pumping on the discharge of a pipe, then you find that temp. had no significant effect on magnitude of water discharge from that pipe, but pumping was highly significant, and you give the discharge of each pumping pressure. These results are ok without any doubt. Go ahead and write your findings.
Assuming that you have hypotheses for each variable, then rejecting a hypothesis is as important as supporting one.
You can. Even though none of your variables are significant you can still write a manuscript. Variables not being significant is also a valid finding.
Vaishnavi Balaji Although it is not as easy to defend such a result as it sounds in some of the replies, I agree with Prof. Ronán Michael Conroy as selection of your variables, method and procedures, and quality of your data are the most important aspects in this regard. If a model was developed based on a solid theoretical foundation and other standards were all up to the mark then sure it can make sense to defend even non significant relationships through logic and specific context of the target population. Tq.
Thank you @Ronán Michael Conroy. I have choosed the variables based on Theory and Theoritical Model.
- Badges
- Science topic
- Similar topics
- Linguistics
- Composition Studies
- Writing