Please enable JavaScript.
Coggle requires JavaScript to display documents.
graduation tests (degrees of freedom (if groups are mutually exclusive/…
graduation tests
degrees of freedom
if groups are mutually exclusive/ form part of a larger set minus 1
start with # groups
minus 1 for every estimated parameter
chi-square test
weaknesses
large deviations are offset by many small deviations(one-tailed test)
Standardized deviations test
Rationale: zx 's should resemble the N(0,1) distributions
a test by eye includes assessing the overall shape, outliers, symmetry
purpose: (Test overall goodness of fit)detect a number of excessively large deviations
small consistent deviations not detected (Tailed Tests)
signs test
purpose: test overall bias.
Rationale: two tailed test, if the the graduated rates are neither higher nor lower than crude rates, then appr half should have +ve and half - ve.
Method: determine the p-value*2, also for m large enough, 20, use a normal approximation
weakness: looking at signs of deviations does not speak to the extent of deviation.
cumulative deviations test
determine statistic, calculate pvalue or determine critical value
purpose: overall goodness of fit; detects overall bias of the same sign
groups of consecutive ages over which the test test is biased
grouping of signs test
Rationale: # groups of deviations of the same sign and compares this with the number expected if signs were arranged randomly G=#groups of +ve , n1= #+ve's, n2=-ve
Reject if: G<= K (Table pg 189)
Purpose: detects clumping of deviations of the same sign
hypothesis: deviations are arranged randomly
weakness: my lead to different conclusions depending on whether we choose +ve or -ve groups
serial correlations test
Rationale: if the is neither undergraduation nor overgraduaton, we expect deviations at individual ages to be independent
rj +ve and large => overgraduation
Purpose: to detect clumping deviations of the same sign.
weakness: correlations for one part could cancel out the correlations of another part