# Thread: Doubt on statistical test

1. ## Doubt on statistical test

Hi,

my perception test presents one stimulus at a time and the listener has to choose one from three possible answers. There are 5 stimuli randomly presented 5 times each. What kind of test can indicate statistically significant responses?

Thank you very much

sid2016

2. ## Re: Doubt on statistical test

What is your research question and hypothesis?

3. ## Re: Doubt on statistical test

Research question: does phonetic context and speech rate affect the perception of X?

Hypothesis: speech rate and phonetic context will have a significant effect on the perception of X.

I have 3 phonetic contexts and 2 speech rates for each context (slow and fast); I want to see which ones favour the perception of X

4. ## Re: Doubt on statistical test

Originally Posted by sid2016
Hypothesis: speech rate and phonetic context will have a significant effect on the perception of X.
Copying and pasting from some old lecture notes of mine here:

Note that a relationship is (or isn’t) statistically significant in the sample
And we use this observation to make inferences about whether or not the null hypothesis is true in the population
Students occasionally get the referents confused. Do not say things like:
“The null hypothesis is that there is no statistically significant relationship between X and Y”
“There was evidence for a statistically significant relationship between X and Y in the population”
The comments above are with respect to a null hypothesis, but same thing for a research hypothesis - your hypothesis is about whether or not a relationship exists (at all) in the population. You use a test of statistical significance to assess the evidence for that hypothesis. Your hypothesis should not be that the effect "is significant".

5. ## The Following User Says Thank You to CowboyBear For This Useful Post:

sid2016 (08-10-2016)

6. ## Re: Doubt on statistical test

OLD HYPOTHESIS: Hypothesis: speech rate and phonetic context will have a significant effect on the perception of X.

NEW Hypothesis: speech rate and phonetic context will have AN EFFECT on the perception of X.

Thanks Matt, is this better as a hypothesis?

7. ## Re: Doubt on statistical test

Sounds better, maybe just describe the direction of the effects too

8. ## The Following User Says Thank You to CowboyBear For This Useful Post:

sid2016 (08-11-2016)

Got it! : )

10. ## Re: Doubt on statistical test

What about the stats to test the hypothesis? Would chi square work fine?

The results I'll be collecting look like this:

Factors: rate (fast and slow) and context (three contexts)
Depend variable: the choices made from three possible answers (a, b and c)

Thanks!

11. ## Re: Doubt on statistical test

Hi,
aren't you interested in the errors made? I mean one of the choices is correct, the other two are wrong, right? So, one way to analyse this would be to look at the number of errors made at each factor setting. If you have several test subjects, you might want to include them in the analysis too.

regards

12. ## Re: Doubt on statistical test

Yes, I'd be interested, but what would the stats test be if I include both right and wrong answers?

13. ## Re: Doubt on statistical test

Originally Posted by sid2016
Yes, I'd be interested, but what would the stats test be if I include both right and wrong answers?
You could use the chi-squared. The null hypothesis would be that the number of errors is the same for all settings, the alternaive that some settings have a different number of errors than other settings.
regards

14. ## Re: Doubt on statistical test

Can I still use chi squared when the factors are: context (three types) and speech rate (fast and slow)?

15. ## Re: Doubt on statistical test

Yes, you will have 6 cells each eith the respective number of errors.
regards

16. ## Re: Doubt on statistical test

Just like this?

Context 1 slow
Context 2 slow
Context 3 slow
Context 1 fast
Context 2 fast
Context 3 fast

But then how can I find out whether factors rate and context are significant or interact?

Also, each cell's expected errors will be 5 or 10...still thinking about it...is 5 enough? Each one of the 6 items will be presented 5 times (or 10).

17. ## Re: Doubt on statistical test

Hi,
yes, that would work. You should have at least 5 in each cell, so 10 presentations look better.

regards