How can we choose the appropriate significance level to interpret based on sample sizes?

[Phuc Nguyen]

With the large sample size, we normally somehow interpret the coefficient significant at 10%, but in the small sample size, we normally ignore the 10% significant level. The reason is that with the large sample size, 10% significant level still makes a substantial contribution.

I am wondering is there any reference or justification about the sample sizes and significance level being noticed? For example, under 1000 observations we just focus on p-value<0.05, or over 100,000 observations, we can focus on p-value <0.1 (just my example to clarify my question).

Much appreciated.

[Carlo Lazzaro]

Phuc:
I do not recall any reading that proposes your first statement (but it may well be my fault).
The issue is that, with large sample size, even a hiccup more or less can make a statistical significant difference. However, ths difference might not have any meaningful effect on the variable you're measuring.
That's why p-value totemism is a matter of endless debate/criticism.
Googling with the key words -ASA p-value- gives back some interesting entries.

[Phuc Nguyen]

Phuc:
I do not recall any reading that proposes your first statement (but it may well be my fault).
The issue is that, with large sample size, even a hiccup more or less can make a statistical significant difference. However, ths difference might not have any meaningful effect on the variable you're measuring.
That's why p-value totemism is a matter of endless debate/criticism.
Googling with the key words -ASA p-value- gives back some interesting entries.

Thanks Carlo, I did read it, a very interesting document