5 Hypothesis Testing Flashcards
What is a hypothesis?
🔸 H0?
🔸 H1?
It’s a prediction about data; that is much more specific than the research question – the experiment design is based on this.
🔸 formulate one so we can test it
〰〰〰〰〰〰〰〰〰〰〰
H0 ➖ The default answer – there is no relationship between variables.
🔸 looking to disprove - not accept
🔸 reject or ‘fail to reject’
〰〰〰〰〰〰〰〰〰〰〰
H1 ➖ there will be a relationship between variables – positive or negative.
What does the P value describe?
🔸 what is the value for statistically significant results?
The probability of the results being down to chance. It tells us whether results are random or not.
🔸 anything UNDER 0.05 is significant
Aside from describing the spread of data, what does SD also tell us?
Can describe confidence levels
What is meant by two/one tailed?
The ends of distributions are called tails.
If we are predicting that the overlap between these tails will be one sides, our predictions are one – tailed.
But if overlap occurs at both sides (high and low scores), our prediction is two-tailed.
Type 1 error
Known as false positive.
🔸 say it’s significant when it’s not
🔸 true but acc false..
Type 2 error
Known as false negative
🔸say results random when acc significant
🔸say false but turns out true
Most common in clinic
Power?
Refers to how ‘overpowered’ or ‘underpowered’ a sample size is.
too large -> overpowered
🔸significant results may be small
🔸statistically significant but not clinically
Too small -> underpowered
🔸not enough examples to reliably tell whether a finding is random or not
What are effect sizes?
The magnitude of difference between groups - degree of overlaps in data.
How do we measure effect sizes?
Cohen's d 〰〰〰〰〰〰〰〰〰〰〰 d= diff btwn 2 means ➗ SD of data 〰〰〰〰〰〰〰〰〰〰〰 EFFECT SIZES
Small 0.2
Medium 0.5
Large 0.6
Larger = bigger difference
Will smaller studies have a larger or smaller effect size?
Larger as sample sizes can effect the results greatly.
