100% satisfaction guarantee Immediately available after payment Both online and in PDF No strings attached 4.2 TrustPilot
logo-home
Summary

Summary Cheatsheet inferential statistics test 1. Grade: 9.

Rating
-
Sold
1
Pages
2
Uploaded on
09-09-2025
Written in
2024/2025

Full cheatsheet inferential statistics test 1. Grade 9.

Institution
Course








Whoops! We can’t load your doc right now. Try again or contact support.

Written for

Institution
Study
Course

Document information

Uploaded on
September 9, 2025
Number of pages
2
Written in
2024/2025
Type
Summary

Subjects

Content preview

Unit 540 and 541 Tests in Rstudio
In green, those are tests for association, you choose depending on the type of Proportion test/binominal test: table (data_name$x), Binom.test(n_yes,total) Goodness of
variable: the fit test: Exp() Obs() Chisquare(,)
For the following codes assume that the dataset is called data_name, change it accordingly
If at least one variable is nominal: Cramer
dependent variable is called y, change it accordingly independent variable is called x
If both are scale and there is linearity: Pearson’s One sample t.test: We are comparing our sample mean with the population mean for
If both are scale and there is no linearity: Spearman’s example 6.5
In purple, those are tests for regression, so when you measure the effect of one or t.test(data_name$y, mu = 6.5)
more independent variable on the dependent variable: - compare CI with the average not with 0
If only one independent variable: Simple regression Paired samples: Assume we have 2 paired samples ex: Results on exam and retake
If more than one independent variable: Multiple regression 1st : Compute the differences in the right order (after – before)
diff = data_name$retake - data_name$exam
In Blue are those tests when the dependent variable is scale, then we choose
2nd: Do a one sample t-test
according to the amount of groups and whether the conditions/assumptions are t.test(diff)
good or not: ** You can directly compute a paired t.test: t.test(data_name$retake, data_name$exam,
1 group or 2 paired samples -> One sample t-test, 1 sample t-test of difference paired = TRUE)
2 groups -> check assumptions/conditions 2 samples : we measure the difference in reading skills among two teach methods
welch test if not equal variance or something wrong with sample size etc... Welch: t.test(data_name$y~ data_name$x, var.equal = FALSE)
two sample/ independent t-test if equal variance and sample size etc... Two sample: t.test(data_name$y~ data_name$x, var.equal = TRUE)
More than 2 samples:
More than two groups -> -> check assumptions/conditions
Welch anova: oneway.test(data_name$y ~ data_name$x, data = data_name, var.equal =
Welch anova if not equal variance or something wrong with sample size etc... FALSE)
Anova if equal variance, sample sizes are similar, no big differences expected.... Anova: 1. model = lm(data_name$y~ data_name$x, data=data_name) 2. summary(model)
data123$group1 = ifelse(data$group == "Group 1", 1, 0)
data123$group2 = ifelse(data$group == "Group 2", 1, 0)
data123$group3 = ifelse(data$group == "Group 3", 1, 0)
Now run the first and second code again without the new created reference group

Onafhankelijk (independent) = oorzaak, wat je verandert (x-as, horizontaal)
Afhankelijk (dependent) = gevolg, wat je meet (y-as, verticaal)
Oorzaak → Gevolg = Onafhankelijk → Afhankelijk
Empirical rule: 68% zit binnen 1 SD, 95% binnen 2, 99.7% binnen 3 standaardafwijkingen.
Statistic: Een waarde berekend uit een steekproef.
Parameter: Een waarde die hoort bij de hele populatie.
Population proportion (p): Het percentage mensen met een bepaald kenmerk in de hele
populatie.
Sample proportion (p ̂ ): Het percentage in jouw steekproef.
Sample distribution: De verdeling van alle waarden binnen één steekproef.
Sampling distribution: De verdeling van een statistiek over veel steekproeven.
Standard error (SE): Hoeveel een steekproefuitkomst kan schommelen als je het experiment
herhaalt.

Confidence interval (proportion): Het bereik waarin het echte populatiepercentage waarschijnlijk zit.
Margin of error: Hoeveel je steekproefwaarde maximaal kan afwijken van de echte waarde.
Population mean (μ): Het gemiddelde van de hele populatie.
Sample mean (x ̄ ): Het gemiddelde van jouw steekproef.
Sampling distribution of the mean: De verdeling van gemiddelden van heel veel steekproeven.
t-distribution: Een verdeling die lijkt op de normale verdeling, maar gebruik je bij kleine steekproeven.
Null hypothesis (H₀): De aanname dat er géén verschil of effect is.
t-value: Hoeveel het verschil is, vergeleken met de spreiding in de data.
p-value: De kans om dit resultaat (of extremer) te krijgen als H₀ klopt.
% difference: Het procentuele verschil tussen twee waarden.


Cramer’s V: Meet hoe sterk het verband is tussen twee categorische variabelen (0 = geen, 1 = sterk).
Chi-square statistic (χ²): Meet of er verschil is tussen wat je ziet en wat je zou verwachten.
Goodness-of-fit test: Test of de waargenomen verdeling overeenkomt met een verwachte verdeling.
Covariance: Geeft aan of twee variabelen samen stijgen of dalen, maar is moeilijk te interpreteren.
Pearson’s correlation (r): Meet hoe sterk het lineaire verband is tussen twee variabelen.
Spearman’s correlation (ρ): Meet of hogere rangen samengaan; geschikt bij uitschieters of ordinale data.
Non-parametric: Analyse die geen aannames maakt over de verdeling (zoals Spearman).
Linear equation: Formule om iets te voorspellen: y = a + bx
Intercept (a): De waarde van y als x = 0.
Slope (b): Hoeveel y verandert als x 1 eenheid stijgt.
Addition: Extra variabelen toevoegen aan het regressiemodel.
Ordinary least squares (OLS): Methode die de lijn zoekt met de kleinste fout (residu).
Two-sided test: Je test of het effect positief óf negatief kan zijn.
Independent sample t-test: Vergelijkt gemiddelden van twee onafhankelijke groepen.
Welch t-test: Variant van t-test als spreidingen verschillen.
ANOVA: Vergelijkt gemiddelden van 3 of meer groepen.
Welch ANOVA: ANOVA die ook werkt bij ongelijke spreidingen.
R-squared (R²): Hoeveel van de uitkomst het model verklaart (%).
Adjusted R-squared: Gecorrigeerde R² die rekening houdt met aantal variabelen.
F statistic / F-test: Test of het hele model beter is dan toeval.
Linear equations hypothesis
One sample: example: is the knowledge of
students significantly different than 6.6


Two paired sample: example: has the
knowledge improved after the master?


Two independent samples: example: the
average scores differs between dutch and
non-dutch (Yes and No equal variance)




More than 2 samples: example: the average
exam results differs between three groups
$8.98
Get access to the full document:

100% satisfaction guarantee
Immediately available after payment
Both online and in PDF
No strings attached

Get to know the seller
Seller avatar
rvdwolf39

Get to know the seller

Seller avatar
rvdwolf39 Saxion Hogeschool
Follow You need to be logged in order to follow users or courses
Sold
4
Member since
6 year
Number of followers
0
Documents
6
Last sold
2 weeks ago

0.0

0 reviews

5
0
4
0
3
0
2
0
1
0

Recently viewed by you

Why students choose Stuvia

Created by fellow students, verified by reviews

Quality you can trust: written by students who passed their tests and reviewed by others who've used these notes.

Didn't get what you expected? Choose another document

No worries! You can instantly pick a different document that better fits what you're looking for.

Pay as you like, start learning right away

No subscription, no commitments. Pay the way you're used to via credit card and download your PDF document instantly.

Student with book image

“Bought, downloaded, and aced it. It really can be that simple.”

Alisha Student

Frequently asked questions