Help writing hypotheses!

[om0303]

Can anyone suggest what my hypotheses may be please? I'm going around in circles.

The study involves:

You suggested items designed to be part of a new questionnaire measure of statistical anxiety.
By virtue of content/surface validity, 20 items were selected to create this questionnaire.
Students on the L7 FDS module were then invited to take part and complete an online survey made up of three questionnaires:

• The FDS measure of statistical anxiety (our measure)
• The Statistics Anxiety Scale (SAS – Vigil-Colet et al)
• The Trait Anxiety Inventory (from STAI – Spielberger)

They first stated whether they believe they had stats anxiety (self-categorised).

We then had to work out:
• Cronbach’s alpha for all 3 measures
• Average inter-item correlation for FDS
• Concurrent & discriminant validity (i.e., correlations between FDS/SAS/STAI)
• Criterion validity for FDS (i.e., ANOVA – can our questionnaire differentiate between self-categorised stats anxiety)

We are testing the reliability and validity of our new measure.

My original thoughts:

- There will be a positive correlation between FDS and SAS (SAS is a similar measure)
- There will be no relationship between FDS and STAI (STAI measures a different construct)
- Then a hypothesis for the anova?

Thanks, happy to clarify anything further

 

[AngleWyrm]

- There will be a positive correlation between FDS and SAS (SAS is a similar measure)
- There will be no relationship between FDS and STAI (STAI measures a different construct)

Two hypotheses, stated as the opinion hypothesis vs it's alternative:

H1: There is a positive correlation between FDS and SAS
H0: There isn't a positive correlation between FDS and SAS

H1: There isn't a correlation between FDS and STAI
H0: There is a correlation between FDS and STAI

The p value of comparison testing gives the level of correlation.

 

[katxt]

H1: There isn't a correlation between FDS and STAI
H0: There is a correlation between FDS and STAI

H0 is hardly a null hypothesis. What result would lead you to reject H0 here? Would p = 0.1 mean there was no correlation?

 

[Miner]

The null hypothesis is typically the status quo (e.g., the treatment has no effect, there is no difference/correlation, the coefficient is zero, etc.), while the alternate hypothesis is the opposite (e.g., the treatment does have an effect, there is a difference/correlation, the coefficient is significant, etc.).

 

[katxt]

Right. You cannot prove statistically that something is zero. If you find no significant correlation, that doesn't mean the correlation is zero. "No significant correlation" is simply a face saving way of saying "We don't know if there is a correlation or not."
You may be able to show that it is close enough to zero as to be of no practical significance. To do that decide before you start how close to zero is close enough (make an indifference interval), and check to see of a suitable confidence interval on the correlation fits inside that interval.
Or, more simply, say H0:the correlation = 0, and when you get a large p value, say there is no evidence for a correlation, but there may be one.

 

[AngleWyrm]

What result would lead you to reject H0 here? Would p = 0.1 mean there was no correlation?

That's a decision threshold you must choose, typically selected as 0.05

For example:

 

[Buckeye]

https://giphy.com/gifs/gKNiW6zL0y0Zl7H1R4

 

[katxt]

H0 is hardly a null hypothesis. What result would lead you to reject H0 here? Would p = 0.1 mean there was no correlation?

A p value of 0.1would mean that that there could well be a correlation, but that the evidence isn't strong enough to claim that there is. H0 can never be accepted as true (because it probably isn't). It just isn't rejected. You just don't know.

 

[AngleWyrm]

There are four possible outcomes to a hypothesis test

 

[Dason]

There are four possible outcomes to a hypothesis test
View attachment 4167

I never really liked the "Accept Ho" terminology. Change that to "Fail to reject Ho" and it's better.

 

[GretaGarbo]

What is this person @AngleWyrm talking about here. It seems very strange.

Two hypotheses, stated as the opinion hypothesis vs it's alternative:

H1: There is a positive correlation between FDS and SAS
H0: There isn't a positive correlation between FDS and SAS

H1: There isn't a correlation between FDS and STAI
H0: There is a correlation between FDS and STAI

And, the p-values does NOT give the level of correlation.

The p value of comparison testing gives the level of correlation.

 

[Dason]

What is this person @AngleWyrm talking about here. It seems very strange.

I've asked myself that before

 

[Buckeye]

With all due respect, I initially thought he/she/it was a b0t. But, I'm not so sure.

 

[AngleWyrm]

correlation and p values

 

[Buckeye]

I meant it in the most sincere way possible. We appreciate any and all contributions.

 

[Karabiner]

We appreciate any and all contributions.

Well...

 

[GretaGarbo]

The p value of comparison testing gives the level of correlation.

A p-value is still not a correlation. And that is not an ad hominem attack.

 

[Dason]

@AngleWyrm There may have been an ad hominem but the concerns are real. And if we're being honest there are times where I feel like you're only contributing negatively to the discussion (because what you're contributing is sometimes objectively wrong and misleading to the OP). We appreciate all the input but sometimes you're just wrong and when people point that out you respond in the weirdest ways and we really aren't sure how to take that.

 

[fed2]

this thread gave me jock itch.

 

[spunky]

@AngleWyrm There may have been an ad hominem but the concerns are real. And if we're being honest there are times where I feel like you're only contributing negatively to the discussion (because what you're contributing is sometimes objectively wrong and misleading to the OP). We appreciate all the input but sometimes you're just wrong and when people point that out you respond in the weirdest ways and we really aren't sure how to take that.

True. It's like that simulation thread all over again. Sometimes what @AngleWyrm says is (being quite generous) ambiguous. Sometimes (like in this case), is simply incorrect.