Please enable JavaScript.
Coggle requires JavaScript to display documents.
JASP - Coggle Diagram
JASP
Effect
Size
the standardized mean difference
ANOVA
Eta squared
Partial Eta Squared
Omega squared
T-test /Non-paramatic alternatives
Paramatic
Cohen's d
Non-paramatic
Rank Biserial
Correlation coefficient
Correlation
Correlation Coefficient r (paramatic?)
Spearman's Rho & Kendall's Tau (non-paramatic)
Regression
Multiple Correlation Coefficient
Odds ratio
2 X 2 table
Phi
Any table size
Cramer's V
Chi-squared
test for independence
Purpose
Relationship b/w 2 or more categorical varaiables
Cross tabulation or contingency table
Findings
Chi square
Chi square contingency correction
Prevent overestimation of significance in small sample size
at least 1 cell < 5
:warning: overly correct - type 2 error
Likely hood ratio
Alternative to Pearson's
Maximum likely hood theory
for small small sample < 30
Nominal measures : effect size
Phi
2x2 table only
Cramer V
2 to 5
range 1 to 0
1 : complete association
0 : no association
Contingency coefficient
large table 5x5
Adjusted Phi
Odds ratio
Reciprocal of odds (1/odds)
times increase or decrease
Post-hoc
Standardized residual (version of Z score)
Z=
(Correction for df =1) Z=
Sign of Z
- Z
score : observed <
expected
+ Z
score :
Observed
> expected : :
Ho : No association b/w two categorical
Compare
Expected if no association
total population
Observed
Validity
<20% cells : < 5
No cells < 1
e.g.
Titanic survivors
Independent variable : column
Age
sex
Class
dependent variable : rows
survival status
Synonyms
Pearson's Chi square test
Chi square test of association
Assumptions
two variables must be categorical
each variable - 2 or more independent categorical groups
Comparing sample to population :silhouettes:
Continuous
Checking for normality -Shapiro-Wilk Test
Non-significant --> Paramatic test
One sample T test
e.g. Height & Weight of sample compare with population
Significant --> Non-paramatic test
Willcoxon signed-rank test
Paired sample test
Ho : Median difference is zero
Differences not follow normal distribution
W < critical value
One sample t test
Ho : Medians of two samples are equal or that the difference between medians is zero
Categorical
Dichotomous category
Binomial
e.g. Laptop - windows/ Mac - UK population & uni students
Multiple categories
Multinomial
e.g. different colour M&M counts in 5 packs
Compared with equal proportion
χ 2 Chi-squared Goodness-of-fit test
Compared with expected proportion
Correlation
Types
Non parametric
Rank based
Spearman Rho
Rho square
Determination
ordinal scale
Kendal Tau
small sample / ties
Assumption violated
Normality
Variance
Parametric
non standardized
correlation coefficient
standardize
Pearson's correlation coefficient
-1 or +1 : high correlation
0 : no
correlation
Coefficient of determination : R square
Explained variation / total variation
0- 100%
0 : no variability around mean
100% all variabily
E.g.
Jump height
Leg power
Features
Not showing the causation
Linear association
Purpose
Association b/w 2 continuous variables
Two groups
Comprarison b/w two groups
t-test, ANOVA
Relationship b/w two variables
Corelation
Regression
Predictive analysis
Predict dependent
From single independent
Simple regression
Multiple independent
Multiple regression
Multi colinearity
Outliers
< 1% of the total participants. The outliers will have a relatively small impact on the model
but keeping them means our sample may better represent the diversity of the population
Linear regression model
Sample size
10 -15 data points per variable
50 + (8 x number of independent variables)
Formula : y = c + b*x + e
y = dependent variable
c = intercept
b = regression coefficient
Comparing
two
independent groups : :beer_mugs:
Independent T test / Student T test
Assumptions not met
Homogeneity - Levene's significant
Welch
adjusted t statistic
Normality
- Shapiro Wilk significant or
Ordinal
Mann Witney U test
Descriptive stratics for non parametric
Median
Rain cloud / Box plot
SD/ SE - confidence interval
Inferential statitics for non paramatic
Effect size
Rank biserial correlation
Location parameter
Hodges Lehmann Estimate to decide Median difference
e.g.
e.g.
Dependent variable
(ordinal) Subjective pain score
e.g.
Grouping variable
in JASP +/- ice therapy - Independent (categorial)
e.g.
2 Groups of independent variable e.g. Males & Females
Continuous dependent variable e.g. 10 weeks post diet weight loss after taking a special diet
Effect size
Cohen's d -
Comparing
two
related groups
Paired sample T test
/ :confetti_ball: dependent T test / Repeated measures T test
Assumptions
Difference b/w groups
Normally distributed
Shapiro Wilk test
No significant outliers
Independent variable
2 categorical / matched groups
Dependent variable
Continuous
compares means b/w two related groups
dependent variable - e.g. weight loss - continuous variable
related group - e.g. pre & post diet
effect size - cohen's d
Non Parametric
Not normally distributed - significant Shapiro Wilk
Ordinal e.g. pre & post hypnotherapy anxiety scores
Wilcoxon's signed rank test
Hodges Lehmann estimate
location parameter
Rank biserial correlation
effect size
Comparing
more than two
means
ANOVA - Analysis of One Sample Variation
Independent ANOVA
e.g.
3 types of Diet
Independent variable - categorical
Weight loss
Dependent variable - continuous
Ho - No significant difference between the means of all the groups
Contrasts - comparisons
Priori test - 6 types
Deviation
simple
Difference
Helmert
Repeated
Polynomial
Post Hoc tests
Identify which groups are different
4 Types
Standard
Games - Howell
Dunnett's
Dunn
4 Corrections
Bonferroni
Holm
Tukey
Scheffe
Sidak
Effect size
Eta squared
Partial Eta squared
Omega squared
Non parametric -
Kruskal Wallis test
Parametric assumptions fail / dependent variable - ordinal
Rank based
e.g.
Subjective pain score
Dependent variable
treatment - compression & cryotherapy, cryotherapy, control
Independent variable
not specify which is significant
Dunn's Post-hoc test
