Use LEFT and RIGHT arrow keys to navigate between flashcards;
Use UP and DOWN arrow keys to flip the card;
H to show hint;
A reads text to speech;
22 Cards in this Set
- Front
- Back
|
Poission Distribution |
Log linear therefore no negative values |
|
|
Assumption of data in Poisson distribution |
Small counts |
|
|
One parameter distribution specifird by the mean ( log linear, Poission) |
Variance=mean Therefore variance/mean=1 |
|
|
Infinity poisson distribution? |
Mathematically identical to normal distribution |
|
|
How to write a log linear function |
Y=e(mx+c) |
|
|
What does a link function do? |
Other residual distributions, using likelihood estimation |
|
|
What is maxium likelihood? |
ML instead of Least of squares |
|
|
Why isnt LS appropriate for non normal errors? |
Not for non normal distribution. LS cause biased estimates for non normal data where ML isn't as its based in Likelihood estimation |
|
|
What do we assume for binomial, Poisson etc in terms of errors |
We use deviances not sum of sqaures |
|
|
What are is Deviances? |
|
|
|
Assumptions about Poisson errors |
Probability of the event occuring is random unlike winning a fight or not However overall distribution might be clumped of aggregated, not completely random |
|
|
What is HF, scale parameter, Dispersion parameter |
|
|
|
Overdispersion |
Common- influenced by response variable that isn't measured, or non-poisson distribution |
|
|
What type of errors can overdispersion cause? |
H.F>1.5 cause for worry! Lead to type 1 errors |
|
|
What are type 1 and 2 errors? |
|
|
|
How to fix Overdispersion? |
Assumed value of scale parameter =1 to actual value, =residual deviance/d.f. and using F-test |
|
|
What do if you cant fix overdispersion |
Assume negative binomal error stucture Tranform or use non parametric analyses |
|
|
Underdispersion |
Less common Type 2 Errors Dealt with Rescaling |
|
|
How to check for H.F in an anova table? |
|
|
|
Difference between negative binomial and poission |
|
|
|
When is negative binomal distribution more appropriate than poisson |
When distribution is aggregated |
|
|
How to describe negative binomal distribution? |
|
