# What does the P-value need to be in order for there to be a ‘trend’ in the data?

Home Forums Methodspace discussion What does the P-value need to be in order for there to be a ‘trend’ in the data?

Viewing 5 posts - 1 through 5 (of 5 total)
• Author
Posts
• #1023
Graham P
Member

Hi everyone.

This is my first post on here so apologies in advance if this is in the wrong section, or just a silly question. Also, thank you in advance for any replies, they’re much appreciated!

I’ve just received a bit of stats coursework back, and I’m a bit confused about one of the comments that the marker has given.

In the results section, I performed a paried t-test and mentioned that there was a trend for one of the dependent variables to be higher than the other, but I also said that this was not statistically significant, and stated p>.05.

However, the marker has commented that there is “only a trend if P is between 0.05 and 0.07”.

I have never heard of this, but if this is true then I shall know for next time. It just seems as though this boundary for there to be a trend is quite narrow. I just assumed a ‘trend’ was when, looking at the data, it just looks as though there should be a difference, but due to things like small sample sizes it isn’t statistically significant.

So could anybody please shed some light on this?

Thank you again!!

Graham

#1027
Dave Collingridge
Participant

Graham,

In all my years of teaching and doing statistics in research I have never heard of a cutoff for stating when a non-significant p-value approaches statistical significance. I think the issue of whether a nonsig p-value trends or approaches significance is a matter of opinion and must be made with effect size and sample size taken into consideration. There may be some misunderstanding between you and the marker. My advice is to sit down with the marker or instructor, politely ask for an explanation, try to reach a common ground on what you meant and what the marker meant, and treat the conversation as a learning experience. You chances for a positive outcome are much better than if you go in with both guns blazing ðŸ˜‰ Â  Â

#1026
Stephen Gorard
Participant

Dear Graham,

Either you have got this wrong or if your quoatation is accurate then the marker has made a serious error. It does not make sense at all.

Â

Perhaps more importantly you need to think about what these p-values actually mean. They cannot help you decide and could be terribly misleading. See attached – a brief extract fromÂ my paper:

Gorard, S. (2014) The widespread abuse of statistics by researchers: what is the problem and what is the ethical way forward?, Psychology of Education Review, 38, 1, 3-10

but this has been well known for a long time (at least the 1930s).

Â

PS you are very unlikely to have trend data and a complete random sample anyway.

#1025

People become very creative when it comes to p-values that fail to reach the conventional treshold of 0.05. I strongly assume the person accounting for the quote wanted to make a somewhat strong argument in the absence of p<.05. But Stephen is right, it just seems wrong.

#1024
Graham P
Member

Thank you for everyone who replied. Sorry I haven’t replied since now.Â

Viewing 5 posts - 1 through 5 (of 5 total)
• You must be logged in to reply to this topic.