Please enable JavaScript.

Coggle requires JavaScript to display documents.

ABM (ANOVA (One-way ANOVA (CRD Completely Randomized Design (Helps each…

- - - - Helps each group get subjects of similiar compositions
        
        SSE
        
        SST
        
        Measures the between groups differences / variations
        
        Large = not consistent with H0
        
        Small = cosistent with H0
        
        Within all groups variation
        
        variation because of nuisance factors
    - - Multiple T-tests
        
        increased chance of Type 1 error
        
        ANOVA stay at 5% type 1 error
    - - Treatment levels AND
        blocks
        
        random assignment
        
        blocking criteria
        
        big impact
        
        not of focal interest
        
        purpose
        
        control nuisance factors
        
        find differences easier
        
        1 more item...
        
        homogenous
  - - - Analyzes variances to find differences
  - - - LSD-Bonferroni
        
        Turkey method
        
        if number of comparisons is large
        
        Decreased type 1 error
        
        2 or 3 comparisons
      - when we tolerate high type 1 error
- - - - Population variances are known
    - - Population variances unknown
        
        Uses weighed average or pooled variance estimator
    - - Variances unknown
        
        Assume population variances are unequal
        
        No pooled variance estimator
    - - Paired or matched samples
        
        Same conditions for both tests
        
        Samples are not independent
        
        T test not good anymore
        
        Purpose
        
        Control for nuisance factors' influences on the outcome variable
        
        Differences can be hidden due to noise
        
        Higher probability of type 2 error
  - - - Use F test when only given sample variances #
- - - - Develop model
        
        Determine regression equation
        
        Standart error of estimate
        
        T test of the slope estimate
        
        No linear relationship = horizontal regression line
        
        H0: B1 = 0
        
        R^2
        
        SSE = unexplained variation
        
        1 more item...
        
        Estimating intervals
        
        2 more items...
        
        Error variable ɛ
        
        captures nuisance factors
        
        indicates inacuracy or error
        
        residuals
        
        differences between actual and model Y values
      - mean of ɛ = 0
        
        mean of ɛ independant of x
        
        ɛi-s are independant
        
        Homoskedasticity: constant variance
        
        ɛ normally distibuted
    - - Pearson
        
        measures direction and strength of linear relationship between two variables
        
        interval of [-1,1]
      - OLS
        
        shows existence of linear relationship
        
        direction
        
        measures one unit changes
  - - - other relevant factors being equal
      - variation of Y better explained
        
        more explanatory factors
        
        Zero conditional mean assumption
        
        can fail
        
        ommited explanatory variables
        
        functional relationship between two variables incorrect
        
        R^2
        
        the more explanatory variable the higher
        
        problem:
        
        even if unrelated or bad variable
        
        Adjusted R^2
        
        strictly decreases with more variables
        
        "penalizes" more variables
        
        1 more item...
        
        multicollinearity
        
        two or more variables are correlated
        
        1 more item...
- - - - compare expected frequencies to observed
        
        expected freq. min 5
        
        larger sample the better
    - - at least 4 intervals
- - - - Similiar to chi squared
        
        Non negative