Understanding the null hypothesis

[Tyto alba]

I was reading Bio-statistics principles and practice by Antonisamy and stumbled upon the following:

Null hypothesis is a hypothesis that suggests an absence of difference, association or effect, the negation of which provides evidence for presence of difference, association or effect.

The only problem I'm facing with this definition is that it doesn't say-

• differences between what,
• association between what and
• effect of what on what?

From random sources I know that a null hypothesis involve hypothesising that a theoretical distribution is consistent with an observed distribution, i.e. there is no difference. Also it can assume that there's no difference between the scores of two variables.

But I don't clearly understand what the book meant by association and effect?

I actually found this post in a different community where the OP is trying to establish a relationship (association) between driving speed and gender. Does the definition refer to such associations?

 

[Dale]

Null hypothesis is a hypothesis that suggests an absence of difference, association or effect, the negation of which provides evidence for presence of difference, association or effect.

The only problem I'm facing with this definition is that it doesn't say-

• differences between what,
• association between what and
• effect of what on what?

Differences in this context refers to differences between the probability distributions of two or more random variables. Since the random variables are usually modeled as belonging to some parameterized probability distribution, this is equivalent to looking for differences in the parameters.

Association or effect just means some function that relates one random variable to another (or the parameters of one to the parameters of another). I don't think there is a real distinction between the two. Maybe "effect" is more associated with continuous random variables and "association" with categorical ones.

 

[Tyto alba]

Differences in this context refers to differences between the probability distributions of two or more random variables. Since the random variables are usually modeled as belonging to some parameterized probability distribution, this is equivalent to looking for differences in the parameters.

Association or effect just means some function that relates one random variable to another (or the parameters of one to the parameters of another). I don't think there is a real distinction between the two. Maybe "effect" is more associated with continuous random variables and "association" with categorical ones.

Oh thank you, that was a huge help.