Please enable JavaScript.
Coggle requires JavaScript to display documents.
Preliminary analysis, t-test, ANOVA (Fitting Models to data (If no diff…
Preliminary analysis, t-test, ANOVA
Preliminary Analyses
Data screening- missing data, data accuracy
Descriptive stat- M, SD, correlation matrix, Reliability coefficients
Assumption testing
general- normality, outliers
specific- multicollinearity, homoscedasticity
pilot test, manipulation check
Assumption violated
-
transform data set - log, square root, reciprocal, reverse score
-
-
Test statistics
varience explained by the model divided by the varience that the model can't explained (effect over error)
expected difference btw population means is always zero (no munipulation, no effect, null hypo is true)
-
-
Fitting Models to data
If no diff btw groups, H0 model fit well
-
If thr is diff, H0 model fit poorly, H1 fit well
-
-
Fit quantified by test statistics - larger, effect present model fit well
Underlying logic
come from same underlying population? - roughly equal, 'no effect' model fit
from diff population - group means should differ, 'effect present' model will fit (H0 rejected)
-
t-test
larger values (>1) mean most of the variation is btw the two group means rather than btw individuals
dependent t-test
Info needed: scores for each participants at Level 1 & 2 (no missing), difference of scores btw Level 1 & 2
-
-
Assumptions
-
Normally disrtibuted difference scores - histogram, Kolgomorov-Smirnov
-
-
-
Independent t-test
VS dependent t-test - diff in each pair of scores (for each participants) {dependent}; diff btw group means (independent)
Assumptions
both groups independent - diff participants, randomly assigned
DV is normally distributed - Histogram, Kolmogorov-Smirnov
Both groups have equal variance - Levene's test, if group size euql, not a problem
-
Underlying Concepts
-
-
if the test statistics is larger than the critical value of t, then the manipulation is effective
-
-
-