research methods

what is an EV compared to a confound

an EV is an uncontrolled factor that affects the DV, the confound is an extraneous variable that varies systematically with the IV to affect the DV

what are the categorical data types

nominal - categorical data with no hierarchical order

ordinal has an order

what are the types of numerical data

interval which is scalar, continuous and has no meaningful 0 whereas ratio does

what are descriptive statistics

give an overall view of data patterns without having to use all the raw data

what is a Z score

a standardised score that is used for comparing a data point to a mean, good for comparison between ppts or conditions

why do we need normal distribution

to be able to predict the population mean from the sample mean and therfore form generalisable conclusions

what are confidence intervals

they show the spread of data and we can be sure the mean of the population is likely in the middle 95%

what is the idea behind sampling error

that ideally we want to get the mean of the overall whole target population but due to practical reasons we cant so instead we have to take a sample from that population of interest.

the sampling error is the difference between the sample mean and the target pop. mean as inevitably they wont be the same

how to reduce sampling error

get an average mean from multiple samples

use larger samples as they are more likely to be better representations of the whole population

describe type 1 and type 2 errors

type 1 (false positive) = saying that the IV has an affect on the DV when actually it doesnt

type 2 (false negative) = saying the IV doesnt impact DV when it actually does

why do we need hypothesis testing

to see whether, if it actually happened that the IV had no affect on DV, so the null hypothesis was true, how likely is it that you get the same results (type 1 error)

what is the alpha level

the threshold for significance at 0.05, if p value is below this then it is significant

it is saying there is a less than 5% chance that the results were found due to chance and not truly the IV affecting DV.

what are the between subject tests

one sample and independent sample t-test

what are t tests for

testing how likely we are to get the same DIFFERENCE between conditions if the null hypothesis is true

t statistic = variance explained by IV/ unexplained variance

what is the shapiro wilk test for

testing for normality - how likely are we to get the same distribution of data by chance

we want a non significant p value

what does it mean if the p value for shapiro wilk is below 0.05

it means that the data distribution is significantly different to the normal distribution expected by chance

describe one sample t tests (assumptions + short def)

assumptions:

• independent data

• continuous data

• normally distributed

data from experiement is compared to a single number

only one DV

what test do we use if one sample t test assumptions are violated ?

one sample wilcoxon signed rank test

describe an independent samples t test (assumptions + if violated + DV and IV no.)

assumptions:

• independent data

• continuous data

• normally distributed

• N of atleast 12

• homogeniety of variance (no significant difference between variance in cond 1 + 2)

One DV

One IV but 2 levels

if violated = Mann whitney-U test

how do we look at homoegeneity of variance

first look at descriptive statistics

then look at visual representations like graphs, histograms eg are they the same shape

to be sure, then look conduct Levenes test ( we want non-sig outcome)

in a Q-Q plot the data should fall along the diagonal line if homogenous

what happens if you fail homogeniety of variance test

use the data in the table from the coulmn, variances not assumed

what are pros and cons of within subject design

pros = time and cost effective

accounts for individual differences

cons = fatigue can cause drop outs esp for longitudinal studies

order effects are likely

what does a matched pairs test look at

the probability of whether the difference between the means in 2 conditions can be found by chance

what are the assumptions for a paired samples t test and what happens if you violate them

• continuous data

• N of atleast 12

• the DIFFERENCES are normally distributed

if violated = use wilcoxon signed ranks test

what does a binomial test test for

the probability of finding the observed proportion given the expected proportion

ie is the observed proportion different from the expected proportion

binomial test assumptions

• nominal data

• scores from a random sample

• independent data

• single dichotomy- one variable two possible outcomes

• you know the expected distribution of scores

what does chi2 test of independence measure

whether proportions are different between groups

assumptions for chi2 test of independence

• nominal data

• randomly samples

• expected N of atleast 5

• independent data

• two dichotomies

• sample size of atleast 40

how to calculate the expected frequency

marginal totals/ N

do this for each box (row total X column total) / N

these should all be above 5

what is chi2 equation

(O-E)2 / E

what test do you use if chi2 of independence is violated

fishers exact test

what is the chi2 goodness of fit test

similar to binomial and requires the steps of chi2 but had more than 2 options

its purpose is to see: Are the observed proportions different to the expected frequency

chi2 goodness of fit test assumptions

• nominal data

• independent data

• randomly sampled

• N of atleast 5 for each category

• one DV with multiple levels

what does a correlation show us

relationships between two variables, two variables are correlated as they both have a relationship with a third variable

correlation does not imply causation though as we dont know what variable comes first

give some Q examples

Qwhat is the relationship between binge drinking and academic success

Q relationship between extraversion and wellbeing

Q relationship between reading ability and IQ

how would you measure the Qs

1. avg, units consumed per session and final exam results

2. score on personality test (measures extraversion +others) and wellbeing questionnare

3. number of errors on national adult reading test and full scale IQ test

what do positive and negative correlations mean

positive = both variables increase/decrease with eachother

negative= as one variable goes up the other goes down

how can we determine the strength of a correlation

use the pearsons product moment correlation coefficient, or pearsons r (-1 to 1)

tells us both the strength of the relationship, if its significant and direction

+- 1 = small correlation

+-3 = medium

+-5 = strong

how can we ensure a correlation is significant

we can give it a p value, relating to the null hypothesis that r= 0 (no relationship)

we are measuring if the correlation is significantly different to 0

if below 0.05 = reject null

how could we see relationships between correlations ?

matrices

what are the assumptions for correlations

• normally distrubuted variables (if not = nonparametric or transform data)

• linear relationship

what are the non parametric methods for correlation

spearmans correlation coefficient (large data sets)

• spearmans rho

• for non normally distributed data (no normality assumption)

• rank each data set and then correlate ranks

Kendalls tau (smaller data sets)