Lecture 13: The Null Hypothesis And P-values Flashcards
Confidence intervals are used to draw conclusions about the ______ ____
Effect size
What’d do we mean by the effect size?
- The size
- The direction
Of the population mean difference
Sometimes researchers are less concerned about estimating the ____ of the difference and instead want to draw a _____
Size
Conclusion
What conclusion do research’s want to draw about the population mean difference?
If there is a population mean difference at all!
(Whether the effect size is non-zero i.e. whether there is an effect!)
How do we make an inference about effect size?
Using a confidence interval! Simply by looking whether the interval includes zero or not
If the effect size is zero then the null hypothesis is TRUE or FALSE?
TRUE! - null hypothesis = no effect
If the effect size is nonzero then the null hypothesis is TRUE or FALSE
FALSE - the alternative hypothesis is true! there IS an effect!
Alternative hypothesis
There is an effect from the IV on the DV
Is the effect size a parameter?
YES - we don’t know its exact value since we can only estimate from a sample = we don’t know whether the null hypothesis is true
Can we reject the null hypothesis?
Only at a given confidence level eg. If we reject the N0 only when the 95% doesn’t include 0, we’ll only reject it 5% of the time when the null hypothesis is actually true
Instead of looking at the confidence interval another way of determining whether the null hypothesis is “rejected” at the given CI is to use a:
P-value
What is a p-value
- A statistic between 0-1 representing our uncertainty about whether there’s an effect
The lower the p-value the more/less standard errors away from zero/1 the estimate effect size is so the stronger/weaker the evidence that there is an effect
- More
- Zero
- Stronger
What does a p-value quantify?
How unusual it would be to get the data we got in our sample by chance if there weren’t actually an effect in the population
P in pvalue is what the probability would be of:
Getting a difference as big as we got in our sample (or bigger) if there weren’t actually an effect in the population
Coin flip example: Say we flip a coin and it comes up heads. If there weren’t no effect, what would be the probability of getting that result by chance?
1/2
It doesn’t mean theres 50% chance the coin is fair it means that the coin IS fair and has a 50% of getting heads
For the coin example, the p-value just tells us how ______ it would be to get the data we observed if the coin WEREN’T biased
Unusual
A low p-value means in the coin example:
- Something unusual happened
- The coin IS biased
BUT we can’t say which with 100% confidence
