How do you know when your data aren’t “close enough”? …and hand grenades!
Null hypothesis – statement that illustrates that there is no relationship between two measured variables i.e. “nothing’s going to happen/change” Ex. “this coin is fair” (will create equal heads and tails) Ex. “ the genes are not linked” (will show independent assortment)
There is a relationship between these to variables… Examples: This coin is NOT fair… (produces more heads than tails) These genes ARE linked… (are not assorting independently)
Are differences between observed (collected) data and the predicted outcomes significant? Our threshold? 5% If >5% chance that random chance caused there to be a difference between the observed and expected = accept the null hypothesis = accept that differences between prediction and observed results aren’t significant (It’s a pretty generous tolerance!)
(Sum of)
CategoryOEO – E(O – E) 2 E Dom, dom12 Dom, rec9 Rec, dom8 Rec, rec.0 Total:∑
How many possible outcomes are there in your experiment? In genetics, this might be phenotypes… Degrees of freedom (df) = # outcomes – 1…
Say our df = 3 and our Chi-squared is 2.96
Probability (p-value) would be 0.5<p<0.3
The difference between observed and expected is Non-significant = accept null hypothesis (~40% chance that variation due to chance events)
CategoryOEO – E(O – E) 2 E Purple stem, jagged leaf Purple stem, smooth leaf Green stem, jagged leaf Green stem, smooth leaf Total: 29∑ = 6.4 x2x2
df = 3 X 2 = 6.4 P = 0.5<P<0.1 Accept the null hypothesis!
CategoryOEO – E(O – E) 2 E Purple stem, jagged leaf Purple stem, smooth leaf Green stem, jagged leaf Green stem, smooth leaf Total: 800∑ = 2.027
df = 3 X 2 = P = 0.7<P<0.5 Accept the null hypothesis!