Stats Test 2

Let's do this!!!

Reproducibility

Same data with same analytic strategy, reproduce the same study to validate the research

Robustness

Change some aspects of the analytic strategies but with the same data to see if the study is validated or remains similar or is totally false

Replicability

Replicate the same study with new data but with the same analytic strategy

Broader Research Systems

Does not follow Open Science! Praises new ideas and significant/publishable results

Individual Research Actions

Does not follow Open Science! Competition for space in the industry, “p-hacking” and prioritizing new research

Open Science Practice

Manuscript (before study to see if you follow up with it in full!), conduction of replications and pre-registrations (publishers will publish you regardless of result if you follow through with truth to the manuscript)

Experimental Design

Cause and effect from the IV (Two independent groups with different conditions) → DV

Independent Groups T-test

Two groups of distinct individuals in two different conditions of the same IV

Two independent groups for the same IV in a study examining difference between groups is equal to →

t-test

t-Test

examines the difference between two group means

What is the null hypothesis and the research hypothesis for t-test?

Null → mean of one group = mean of the other group

Hypothesis → mean of one groups differs from mean of the other group

Another hypothesis → mean of one groups is higher than the mean of the other group

Does t-test have directionality?

YES

Steps for a t-test:

→ what is the hypothesis (directional or no?)

→ alpha level (normally 0.05)

→ figure out the t-obtained

→ figure out the t-critical

→compare t-obtained to t-critical

t-test obtained checks for what?

variability between / variability within

What does the critica value of T depend on?

→alpha level

→ direction

→ df

Independent group design

→Quasi experimental

→ experimental

Quasi-experimental

Is when there are two independent group with independent individuals from different populations

Experimental

Individuals come from the same population and are randomly assigned into two distinct groups!!

For t-test what do we use to check for effect size?

Cohen’s d

Paired Groups t-test

Same individuals but evaluated in different time frames or with individuals from different groups entering in pairs for comparison

Logic of Paired Groups follows:

Figurering out the difference between the pairs and the standard deviation of the pairs

Does t-test equations change from independent to paired design studies?

Yep, keep an eye on that in the test

Larger sample size makes the shape of the distribution?

• decreases standard error

• reduces variation

• sample means closer to true mean

• more normality

t-distribution

similar to normal distribution but it is dependent on degrees of freedom for its shape. Accounts for extra uncertanty so it decreases the high mean and increases the tails

Assumptions for t-tests

→ Scale of measurement is continuous

→ random sample that is representative

→ homogeneity of variances the variances of independent groups are approximately the same

→ normal distribution is close to the t-distributioin

Check Homogeneity of Variances with? And if it does not follow it which test should you use?

Levene’s test, if p is smaller than the alpha than the assumption has not been followed and it is suggested not to move on with the t-test

you should than use Welch’s !!!

Normality is what?

Shapiro’s test if p is smaller than 0.05 normal distribution is not assumed so the t-test should probably not be conducted

QQ plot checks what?

Normality of the data , the points should be straight if normality is met

If you cannot assume normality you use what type of tests?

non-parametric tests

What does ANOVA stands for?

Analysis of Variance

If you examine the difference between groups but there are more than two groups what do you use?

ANOVA!!

What has levels IV or DV?

IV - categorical

How many IV are in the ANOVA?

Only one, but there can be multiple levels. Like candy bar is the independent variable and kitkat,reeses,hershes are three different levels aka groups

One-way Anova is =

One IV!!

Why can’t we do a bunch of T-test instead of doing an ANOVA?

Conducting multiple t-tests does not account for the accumulation of alpha level, so it increases the Type I error (falsely rejecting the null)

Family-wise alpha

A percentage that accounts for the accumulation of alpha levels during an ANOVA test, the percentage identifies the probability of making a Type I error when perfominf multiple tests

So the alpha level in ANOVA is not 5%?

The thsehold for each group is 0.05 as always but the accumulation of all of them has to be calculated with Family-wise alpha

One-way ANOVA is a type of what test? t-test?

NOPE, F-test

Steps of One-Way ANOVA

→state hypothesis

→ set the alpha level

→ select the appropriate test startistic

→compute the test statistic (obtained value)

How does the null and hypothesis work for One-way anova?

null is that the general groups are all the same for the DV

hypothesis is that the general groups are different but it is not explicit what is the directional of difference from one another !!!

What does larger F values indicate?

greater proportion of variance is likely due to the IV (which is good!! shows that the IV actually makes the change)

F-distribution

Has no direction so F cannot be less than zero!
F ration should be around 1 for H0 = it is the variation that should be 1 because for the H0 there is no variation because they are the same

To find F-critical you use what?

→ alpha level

→ degrees of freedom between

→ degrees of freedom within

What do we use for effect size of anova?

Eta-squared

Eta-squared

A percentage that shows from the DV of 100% in general in this case, how much can we explain with our findings in real life ? The answer is the percentage of the eta-squared

Tukey’s HSD post-hoc test

It is used to identify the directional difference between groups ! To see which groups really have the statistical significant difference

Saying the group means are not the same does indicate that the groups means are all stat sig different. T or F

F!! It does not that is the whole point of Post Hoc test

Pairwise Comparisons

Used in Post-hoc testing to identify the comparison of two means

If the mean difference >= HSD?

The pair means are statistically significant