IPR

Thematic analysis (Braun & Clarke, 2006)

"Thematic analysis is a method for identifying, analysing and reporting patterns (themes) within data. It minimally organizes and describes your data set in (rich) detail." (p79) •No specific theoretical commitment •Generating themes across participants •Themes don't &"emerge" - we identify them •Themes usually more complex than just categories • •E.g. Interviews on disability: "Adapting to change" (which doesn't offer any insight) vs "Feeling like a ghost: loss of identity and social status"

Content analysis (Hsieh & Shannon, 2005)

"qualitative content analysis is defined as a research method for the subjective interpretation of the content of text data through the systematic classification process of coding and identifying themes or patterns" (p1278) Three kinds: •Conventional - don't start with pre-conceived categories, use open-ended data collection •Directed - have some prior categories in mind •Summative - how much are certain terms used or implied? (similar to linguistics)

Importance of variability

-statistics is all about making sense of variability -all humans differ in behaviour, actions, reactions, etc -understanding this variability is key to psychology -this allows us to make the unpredictable predictable -very few statistics are done by hand as most are now done on a computer All statistical tests boil down to the same basic principles which is variability within our control divided y variability outside of our control. For example if you are measuring pleasure and you have two experimental conditions, in group 1 people get chocolate and in group 2 people get an electric shock. you expect that those who receive chocolate will gain more pleasure than those who get the electric shock. However there will be people who don't like chocolate and people who like getting an electric shock. that is the variability that is out of our control. We will always have this variability but just because we can't control it doesn't mean it shouldn't be accounted for.

what is bias?

-the checkerboard problem brain understand how checkerboards work and how shadows work so we perceive the checkerboard tiles as two different colours - this is an example of unconscious bias and heuristics, bias is the prejudiced or unsupported judgement in favour of or against one thing, person or group as compared to another. in a way that's usually considered unfair. heuristics this is the approximate strategies or rules of thumb for decision making and problem solving, that do not guarantee a correct solution but that typically yield a reasonable solution or bring one closer to hand 'heuristics are quite useful but sometimes they lead to severe and systematic errors' Tvertsky and Kahneman, 1973 the initial heuristics were established by Tvertsky and Kahneman. These were representativeness, availability and anchoring and adjustments. representativeness people make decisions based on categories people overestimate the effect if similarities on predicting an event eg stereotyping availability the ease with which an idea can be recalled. if you see something regularly or have seen it recently it is more easily recalled. anchoring and adjustment when individuals depend too heavily on some initial information (the anchor) in making a decision. eg when it says somethings n sale for 50% off, are you really getting a good deal? is it every sold at full price? Schemas mental frameworks that bundle knowledge together in an organised way eg categories. attitudes and stereotyping based on schemas we automatically assign individuals to salient and accessible social categories. Our brain automatically considers information we've come to associate with that group. attitude an association between a category and an evaluate valence (eg a gut feeling) stereotype A specific associating between a category and a specific trait can be implicit or explicit and can be developed through direct contact or through vicarious contact.

Inputting data in Jasp

-you cannot directly input data in to Jasp you have to use Excel or numbers -you additionally have to save the excel/numbers file in to a CSV file to be able to use it -when you open Jasp, click on the three lines and press the open pop up and then computer then select the data file you wish to use -if you double click it should bring you straight back to excel/ numbers. -synch the data with the computer file so that if you have to make changes to the file these changes will automatically appear in JASP as soon as you hit save in the excel/ numbers document -to do formulas in numbers simply go on to insert and select what formula you would like eg mean or sum you can drag down using the yellow box around the cell for the formula to then be done for the rest of the row/ column to add 1 to all the numbers in the column you would click insert, sum, then the number and then add +1. the drag down for the rest of the column you would need to make this in a new column you can paste numbers as 'values only' in order to avoid having the formulas still attached to them

Kahneman Systems 1 and 2

1- fast, unconcious, automatic, everyday, error prone (amygdala) 2- slow, conscious, effortful, complex decisions, reliable

Tips for presentation

1. Consider will a graph actually help the reader or could it be summarized by descriptive statistics 2. Include a title and label all axis. Use APA format. There are certain conventions for psychology we should adhere to. Any graphs are known as figures, you do not need to label them as bar charts as the reader will know 3. Scale y axis properly and proof read labels. Don't alter scale so that the difference between groups on a bar chart looks larger than it is. This is an example of 'cheating by charting'. Some reachers say you must start the y axis as zero, however this is not always appropriate eg with most line graphs. If you do not start the scale at zero you must say that you did this and why. 4. Portray the relationship in the most simple way possible 5. Use the accent principles Apprehension- ability to correctly perceive relationship between variables, does graph maximise understanding of relationships between variables. Clarity- ability to distinguish elements of a graph. Are most important elements or relations visually the most prominent Consistency- ability to interpret graph based on similarities to other graphs. Do the elements appear the same as other graphs? Efficiency- ability to portray a possibly complex relationship in the most simple way possible. Are elements economically used, is it easy to understand? Necessity- is there a need for a graph? Truthfulness- ability to determine the true value represented by any graphical element by its magnitude relative to scale used

Jasp data visualisation

1. Open descriptive stats data 2. Go to descriptives and click on the plots drop box. 3. Click basic plots and go to distribution plots to give you a histogram for group 1. 4. You can add in groups 2 and 3 and the data will stack histograms for groups one on top of the other 5. Click off distribution plots and click on box plots , this stacks them again 6. If you want everything on one row so can directly co0mpre box plots see 'split section under variables. Synch the data, open the file on excel. Copy and paste data from each group on top of each other, then label this coloumn all data next to the all data colmoun write a group coloumn and label the data according to what group its in eg first 50 in group 1, next 50 in group 2 etc. 7. Reopen JASP, go back to descriptives, select all data . then go on to the split below the variables box and select group and press the arrow to split. This will give you all the groups on the same chart in a row. 8. If you want to remove group 2, right click on undesired group on the normal first page. Go on to filter and click the tick of group 2 and then it will be removed from the histogram. 9. Violin element- this is a histogram but smoothed and turned 90 degrees and put on a data box plot. And then mirrored. You can look at violin elements without distribution points , good for data relationship between groups 10.Jitter element - this allows you to see each individual data point. 11.Can change colour palette by clicking on the colour palette drop box and clicking the use colour palette tick box. 12.You will notice on JASP there is no bar chart option. You can do this instead in excel. Bar graphs aren't that useful as they only show you one value really, and can't see the distribution. For example when using the mean you won't see any of the data above it, which is where roughly half of it would be .

The 2 main functions of visualising and presenting data:

1. To help you understand and digest your data before you do statistical analysis 2. To present data effectively to others weather it is in a lab report, poster, presentation, dissertation or submitting to a journal

strength and direction of a linear relationship

2 continuous variables (interval or ratio) -1 to +1 0 is no correlation requirements interval or ratio data (both variables) linear relationship no significant outliers data normally distributed as this is a parametric test 1 tail or 2 tail state hypothesis

correlations introduction

Aim to study causal relationships There is no direct manipulation of the variables this is observing natural events the relationship between two variables focuses on the degree (magnitude) and the direction (positive or negative) you can ask on a scale of 0-100 for two variables. if there is a perfect positive correlation will have the highest number for both values. look at a visual representation- a scatter gram positive correlation: when x increases y increases negative correlation: when x increases y decreases

Bar charts and line graphs

Bar charts and line graphs are used to represent mean scores, typically they are used to compare groups or conditions of a nominal variable (IV) The general convention is for the IV groups to go on the X axis and mean scores to go on the Y axis. Clustered bar charts can be used to display more complex relationships between variables. Multiple line graphs can be used to display complex relationships between variables. Bar chart vs line graph Depends on data type and personal preference Can use bar charts when data comes from nominal groups, theres no logical middle point between groups Line graphs can be quite messy when there's lots of data. Bar charts and line graphys are probably the most used to represent mean data across a group. However scietists have received back lash as bar charts and line graphs don't tell us the whole story. Violin plots are best

Ethics

Ethics and society- Ethics are very clearly apparent in medical/ legal situations. There is often a grey area in ethics, if in doubt doctors will treat even if a patient has refused. Factors such as duty of care, consent, free will must be considered. In the eyes of the law when a person can't make a decision for themselves it is up to their spouse (even if they are separated) and then the next of kin. Autonomy- respect for an individual's wishes, their capacity to make decisions, and the maintenance of their independence Beneficence- does an action make a situation better or worse? does the action plan prevent harm? does the action provide appropriate psychological or medical treatment? Justice- does the action balance the needs of the individual with the needs of society? does action distribute resources fairly and equally? does the action distribute the wishes of the friends, family and individual fairly? Capacity- legal concept. The ability to make a specific decision at a particular time. may be highly specific in terms of time and domain. However a lack of capacity does not mean that an individual's wishes shouldn't be respected. Adults are assumed to have capacity unless stated otherwise Best interest- when an individual doesn't have capacity and action is required. issues arise such as discrimination because sometimes a superficial assessment is made. Also it must be assumed that the individual may regain capacity. The individual must be involved as much as possible. An individual's past, present wishes feelings and values must be taken in to account. Must consult these as widely as possible. Next of kin- legal term. for children under 18 this is a parent, they have a role in making decisions. Power of attorney is when someone is appointed power. BPS- the ethical body that we as psychologists are under GDPR- general data protection regulations for the European union. This relates to personal data from which one can be identified with a fair degree of certainty. There are penalties for breaches and misuse. Individual's have rights to access it such as deletion or correction.

correlation coefficient

H0 is that there is no relationship between the two variables H1 will be that there is a relationship compare the scores and then divide by the sum essentially

Reflexivity

How did the presence data make you feel? Have you had similar experiences? What did you think of the people participating? refexivity is acknowledging that you have on fact made a judgement on that data. even if you think the data is boring, its can effect how closely you've read the data. So you will look at the data in a way that supports your own analysis. Basically the idea of 'check your baggage'- what past experiences are you bringing in with you to this experiment? You can't ever free yourself from qualitative bias. "Reflexivity is commonly viewed as the process of a continual internal dialogue and critical self-evaluation of researcher's positionality as well as active acknowledgement and explicit recognition that this position may affect the research process and outcome." (Berger, 2015)

how to do a bar chart on jasp with 4 bars on 1 chart

In JASP load up data by opening the CSV file. stick data in to descriptives and then select box plots we can see that we get them all stacked on top of each other but we want them all on the same graph so we can have a look. go in to excel and stack all data in one column and label it according to which data set it is from eg sham close, tms close, tms far, tms far, sham far. select all data and split by group. Then go to box plots and they will then all be on the same thing. quite a bit of overlap for far sham and far TMS. for near you can see that error bars don't lap over that much and there are some outliers. so it looks like TMS has effect in near group as oppose to far open excel take the means from Jasp paste it in to excel and then control click to select the separate cells it will paste out on to to make a beater table then paste further down. then highlight the pasted neat area then press insert and select a bar chart. if you want to rearrange the columns you have to rearrange the cells in excel by copying and pasting them to the right positions. if we then also want error bars we can select the bar and then go to add chart element and then click error bars. ca add standard error or other options, but click 'more error options', this allows you to click add custom value, this allows us to tell excel exactly the numbers we want. press control and select the 4 standard deviations from the table. so now graph has those standard deviations in as error bars. This has accidental changed the axis, and made the data look different, makes difference look smaller select the axis and press options and vertical value and change minimum and maximum to the data set. ali used error bars which are smaller than sd bars which is why the bar chart you make may look different.

4 scales of measurement (NOIR)

Nominal- simply names or categories, you can only ask the question ' are these the same or different?' eg are person A's eyes the same colour as person B's? Ordinal- data that can be placed in order, you can ask the questions is it the same or different and is it greater or smaller than something else? eg likert scale Interval- tells you the degree of difference between two scores. you can ask the questions- is it the same? is it greater or smaller? by how much is it greater or smaller? ( Interval has no true zero) eg temperature Ratio- You can ask the questions is it the same? is it greater or smaller? by how much is it greater or smaller? you can also find the ration between numbers and it has a true zero point eg height

Visualising data- distributions

One of the first steps in data analysis is to take a look at the distribution of scores in your data This is important as the distribution of scores in a data set can determine what further statistical analysis we should conduct on that data We will go through histograms, box plots and violin plots X axis along the bottom, Y axis up, 2D plots will always have these Occasionally you will come across 3D plots where a third dimension is usually specified on the z axis. Histograms show the overall shape of distribution scores from a sample The X axis shows the scores of the variable, in a histogram the data is divided in to 'bins' On the Y axis, the counts represent the frequency of the scores which occurred within that range. Most statistical packages automatically put data in to bins for you Can do this in excel Normal distribution In nature the normal (gaussian) distribution occurs incredibly commonly IN psychology the variables we measure tend to follow the same distribution but this is not always the case. Distributions allow us to infer things aboyt population.

qualitative data

Qualitative research is about trying to understand the meaning and feeling of data as oppose to trying to sum it up in a way that reduces it to quantity and numeric concepts. This sort if stuff can't always be quantified. Qualitative research does not always have to be used in isolation from quantitive data and you can combine the two in one paper. "Qualitative researchers tend to be concerned with meaning. That is, they are interested in how people make sense of the world and how they experience events. They aim to understand 'what it is like' to experience particular conditions (e.g. what it means and how it feels to live with chronic illness or to be unemployed) and how people manage certain situations (e.g. how people negotiate family life or relations with work colleagues). Qualitative researchers tend, therefore, to be concerned with the quality and texture of experience, rather than with the identification of cause-effect relationships." (Willig, 2001, Introducing qualitative research in psychology) •Shared principles •Quantitative methods alone can miss nuance, complexity •The researcher contributes to the knowledge that is uncovered ( qualitative data acknowledges that the perceptions of the researcher may affect the interpretation of the data and are thus not objective).

Box plots (box and whisker)

Simple but effective visualisation of the distribution of your data. They are also used to compare across groups. Additional benefit is that box plots can be used to highlight any outliers in the data The outer bounderies of the box plot are simply the minimum and maximum scores. Most statistical packages employ methods of calculation in which overall boundries of the plot are based on the minimum and maximum values excluding outliers. Whiskers give outer ranges. Black line in the middle of the box represents the median score. Upper and lower quartiles are at one end of box each. Both halves of the box together are the inter quartile range. Box plots give you an indication of how your data is distributed. Check data to see why outliers have occurred

Skewness

Skewness measures the symmetry in the distribution. I.e. weather it is skewed to the left of the right Positive skew is when it is bunched to the left Negative skew is when scores are bunched to the right

Abstract

Summarise the experiment in 200 words so 50 words on each section eg aims, methods, results, conclusion, discussion. This should also be written last you need the discussion / what does it mean (the so what part)

wilcoxon 1 sample in JASP

T test- 1 sample t test (the test value is the previous population median) select wilcoxon signed rank select hypothesis binomial probability given need descriptives (median)

Kurtosis

The shape of the distribution Is it too spiked or too flat? Negative kurtosis= flat, also sometimes called platykurtic Positive kurtosis = when it is too spiked so you don't get too much range , also sometimes called leptokurtic

Title

The title should describe experiment completely and appropriately Must include the IV and the DV write it last the title should either summarise the results or pose a question

Types of data

There are 3 main functions of statistical techniques descriptive statistics- provide ways of summarising information Inferential statistics- confidence with which we can generalise from a sample to the entire population. data exploration techniques- makes sense of large amounts of data When you carry out your experiment make sure you use descriptive stats first. You can do this numerically through operations such as mean but most of the time it is done graphically eg a bar chart.

comparing Mann Whitney U 2 independent groups

There are two different conditions with different participants in each condition. Mann whitney U is also sometimes called the Wilcoxon rank sum test. You use it when there is categorical or abnormal data, indepdant measures design, data appropriate to the test, you decide wether its 1 tailed or 2 tailed. you determine the hypothesis H0= 2 populations are equal H1= 2 populations are not equal eg you test 10 students 5 do mindfulness meditation, 5 do focused meditation. the afterwards they rate their stress out f 10. so no difference assumed. what does this mean in terms of rankings. we should observe similar numbers of high and low ranks across conditions. total number of ranks to be the same. rank data in ascending order ( what if 2 are the same? then you have to calculate 'actual rank' this is where you add up the sum of the 2 and then divide by 2. 2. ad up all the ranks separate for the 2 groups. 3. find the smaller sum of the ranks and call that one t this allows you to calculate U

Wilcoxon signed rank test

This is the difference between two dependant groups The nan-parametric equivalent to the paired sample t test It can also be used as an equivalent to the one sampled t test It looks at the frequency of signs so you need: a paired or matched group, appropriate data, to determine if its 1tailed or 2 tailed and the hypothesis so you measure the differences (post-pre) rank the differences affix a sign to the differences remember to ignore the zero's If the groups are equivalent there should be the same amount of high and low scores in conditions if equal you take the null hypothesis, if different take the alternative hypothesis Can now take the wilcoxon value (T). add up all the rankings with a positive value (T+), and add up all the rankings with a negative value (T-) T is the smallest of these two values N is the total number without any zeros Compare T to the critical value if t is less than or equal to the critical value for then N and significance level then you reject the null and accept the experimental W is the T value when sometimes written sometimes you can carry out the sign test on data and not get a significant result but you carry out the wilcox test and you do. This is because Wilcoxon takes in to account the magnitude of differences and not just if they're positive or negative.

comparing one sample to pre-deduced value

This is the non-parametric equivalent to the one sample t test You can either use 1 sample sign test, or 1 sample Wilcoxon signed rank test what you are essentially looking for is a significant difference between the medians hypothesis is different for this test but everything else pretty much remains the same H0= median samples are the same H1- median samples are different sign test for 1 sample you need to know if median value is the same for them all first calculate the difference then affix + or - sign calculate N which is the total number excluding all those with zero/ no difference using JASP input all variables in to a binomial distribution end up with probability If you are doing a two tailed test it is important to bear in mind that Jasp will only give you results for a 1 tail so you need to motile the value given by 2. For the Wilcoxon calculate the difference rank and then affix signs calculate T+ and T-.

Spearman's rho

This measures the strength and direction of a relationship between 2 variables ordinal data for both Spearman's Rho involves a monotonic relationship which means if one variable is needed to decrease to decrease another this is a positive correlation and vice versa

Violin plot

Violin plots have become increasingly used in recent years The main advantage of violin plots is that you can combine what is shown in a histogram with what is shown in a box plot It shows you: the average, range, and expected maximum and minimum scores. And the distribution of scores Violin plots are available in JASP In a violin plot, a stradard box plot is generates and the violin element is laid over the plot The violin element represents the frequency of distribution scores Visualising data between variables (bar charts and line graphs) Mary Eleanor spear pioneered use if the bar chart and box plots in the 50's Also wrote a book about how charts can be used to mislead

non- parametric test

a family of statistical procedures that do not rely on the restrictive assumption of parametric tests. In particular they do not assume the data is not normally distributed you don't know about the population/ data cannot be normally distributed.

parametric tests

a test that requires data from one of the large catalogue of distributions that staticians have described. assumes population is normally distributed. need normally distributed sample, homogeneity of variance interval or ratio data.

how to calculate binomial cumulative probability in JASP

calculate the difference and make a separate column labelled distance This difference should be the post condition - the pre condition You now need to get the median so go on to descriptives then statistics and then median now go in to distributions which should be on the top bar, if it is not here you can click the blue '+' sign and get the distributions. click on 'discreet data' and then binomial click cumulative probability you then get the interval between 0 and the number of successes as your value

Measures of central tendency

central tendency means some kind of average value most common are mean, median and mode the mean is the most commonly used and understood. It is when you add all of the scores and then divide by the number of scores there are. The x with a line over the top is the mean. the x by itself is an individual score sometimes written as x with a subscripted i. The symbol next to the x is the sigma which means the sum of and n is the number of scores. the median is the point on the scale with which there are half of the obervations above the value and half of the observations are below. The middle score on this set of numbers when they're organised in numerical order. mode is the most frequently occurring score in the data set. If the distribution of data follows a perfectly normal distribution, the mean, median, and mode should all be the same. In data at best data can be nearly perfectly normal, at worst it can be very skewed. In very skewed data the means of average will be very different. With really skewed data median can sometimes be considered the better average to use than the mean , as the mean is influenced by extreme values. when distribution is roughly symmetrical however we can use the mean. the mean should also be used for interval or ratio data and should be used when doing a parametric test. the median should be used when using ordinal data and non parametric stat tests.

How to use the Wilcoxon signed rank test in JASP

check for your assumptions in the descriptive statistics create difference column have a look at descriptives and distribution plots ( allows you if you have data that is ratio for example to check that data is truly abnormal so you can determine if this is the right test to do. click on T tests, paired sample, After and before should be placed in variable pairs. select wicoxon signed rank test. put the hypothesis direction. select effect size. P value is small then significant, large it is not significant

chi squared intro

chi squared is a non-parametric test. A non- parametric test is a test that does not assume anything about the data, in particular it does not assume that the data is normally distributed. Unlike parametric tests which do assume this. The use of a non-parametric test suggests we don't know a lot about the data. A non-parametric test measures the difference between the observed and the expected frequencies of the outcomes of a set of variables. It also tests the goodness of fit between an observed distribution and a theoretical distribution of frequencies or to test wether two variables are independent or related to one another. goodness of fit means does a sample match/ represent the data you would find in the actual population. Or you fit categorical data to a distribution. test of association- categorical data from two independent variables to see if there is any relationship between them. Chi squared can be applied to when you have 1 or two categorical variables from a single population. If you only have 1 categorical variable chi squared answers the question of how the observed value is significantly different from the expected value. If you have 2 categorical variables, chi square answers the question if there's a significant association between these two variables.

how to get descriptive stats for 4 sets of data on JASP

click the green + button to add 4 total data sets then select the data set you want for each file eg select 'happy full', 'sad full', 'anger full' and 'scared full' for the full body data set edit titles of each of the folders to what the data set is eg 'happy', 'sad', etc. to get just the standard deviations and means go on to the statistics section and then deselect the ones you don't want such as minimum and maximum To copy all the data sets at once simply click on results and then copy and then paste in to your word or pages document

The sign test

compare the size of two groups uses + and - instead of quantitative numbers. is the equivalent to paired/ one sample t test to be used on one person, abnormal data named or paired DV, categorical data determine if your hypothesis is 1 tailed or two tailed. 4. so hypothesis will be H0 it will be half and half as this is a test for before and after take away the pre score from the post score affix signs to the differences ignore all the data which has no difference/ a difference of zero count the number of +'s and -'s . N- is the total number of samples with a difference (all apart from those with zero) find the P value using binomial distribution probability there is a fixed number of independent trials with only two outcomes X= number of successes n= number of participants for the binomial distribution select binomial, show distribution probability of success is 0.5 with two outcomes suggests no difference The interval should be the smaller of the +'s or -'s. Now compare your P value to your significance level chosen (always 0.5 for yes/no or two outcomes). When the P value is larger than the level of significance we fail to reject the null hypothesis and have to accept the alternative hypothesis.

one sample t test

compares mean of single sample to predetermined population mean eg if you compare a group to the average. use it when you want to compare a single sample mean to a predefined, specific value. difference between mean of population over the standard deviation of the mean. compare group to known value interval or ratio no outliers approx normal distribution one tailed or two tailed null hypothesis is that the sample mean will be the same as the population mean alternative hypothesis is the sample mean will be different from the population mean. from Jasp we get t value, degrees of freedom, p value and Cohen's d (effect size)

How to calculate chi square by hand

create the chi square contingency table. with data in matrix format state your hypothesis/ assumption- H0 is the null hypothesis which states there is no relationship between variables. H1= this is the alternative hypothesis that suggests the variables are dependant in some way, and thus the results are not random. fill the table with the observed values calculate the expected variables. this would be the values that would be observed if there was no relationship between variables. This is shown by the equation E= Mr xMc n E= expected value Mr= row total for cell you are calculating expected value for Mc= column total for that cell n= sample size 5. For each cell you subtract the expected value form the observed value. Then square this number and then divide it by the expected value. You then add together the total for all the values 6. you then calculate the degrees of freedom and the critical value. the degrees of freedom is the number of independent variables that went in to a calculation. Fo chi square this is rows -1 X columns- 1 you the use the chi-square distribution table and find the critical value at the intersection of the degrees of freedom and the level of significance. statistical significance is the probability of rejecting a null hypothesis that was in fact true. At the level of 0.05 there is a 5% risk of concluding that a difference exists when it actually doesn't. The most typical levels used are 0.05 and 0.01. The lower the level of significance the lower the risk of error. chi square only gives a yes or no as to wether there is a relationship it does not point out what accounts for the significance.

how do we do Mann-whitney U on JASP?

dependant variable column, click on and make sure data is labelled as ordinal if ordinal put the independent variable in to the variables window and the dependant variable in to the split. select statistics and look at the median and mode and interquartile range distribution plots to look at the distribution if interested select T tests independant put the IV in to the variables DV in to the sampling click Mann-whitney U tick the correct hypothesis for if you chose one tailed or two tailed.

descriptive vs inferential statistics

descriptive statistics- provide us ways of summarising information from our sample- eg mean, standard deivation, etc inferential statistics- allow us to analyse the confidence with which we can generalise the findings from a sample to the general population modelling data- can give us a logical sense of order and structure to our data

sample vs population

descriptive stats describe the characteristics of sample data measures of central tendency are mean, median and mode measures of variation/ spread are range of scores, inter quartile range and standard deviation. inferential statistics allow us to infer properties of a population based on our sample examples of this are correlations, t tests , chi-square, etc (anything we can generate a p value with) usually a population is too large for a researcher to attempt to study all of its members. On such occasions a smaller carefully selected sample can be used to represent the larger population. we can use the sample to infer properties of the population. However, how can we be sure that this sample is representative of the population? in short we can't really. Inferential statistics allow us to asses a degree of certainty when generalising findings to the population.

chi square results section

do not start with a table. write an introductory sentence such as 'a chi-square test was performed to analyse the relationship between... there is/isn't a significant difference'. You write it out as X2(two should be superscripted here), X2(2[ where 2 is the degrees of freedom], N=741 [ where the number of participants is 741]). The chi square value should be written to 2 decimal places to conform with APA standards. and the p value should be written as <0.01. go back on to JASP, and copy and paste the table in APA standard this is no horizontal lines other than under the title and no vertical lines.

how to do chi square in JASP

elect frequencies at the top grey bar, and then select contingency table. select which variable you want in the rows and which variable you want in the columns by selecting them and using the arrow button to move the variable in to that box. This automatically does the frequency, it also automatically does the chi square value, degrees of freedom, and significance level if the chi square test is open/ selected Jasp also automatically does the expected values as well If you want to see these open 'cells' drop box and select expected and now you can view them.

how to report results for all the T tests

find in IPR lecture 12 pictures

many types of cognitive biases

fundamental attribution error tendency for attributors to underestimate the impact of situational factors and the tendency to overestimate dispositional factors. Ross, 1977 eg assume someone is late to work because they're bad at thing as oppose to traffic was bad. intergroup bias social identity- an individuals self concept derived from perceived membership in relevant social groups. The status of the group is determined by social comparison of the out groups. when the in-group has higher status than the out group this reflects positively on in group member's self concept. individuals want to feel good about themselves so they favour in-group and discriminate out group leading to intergroup bias. halo effect tendency for an impression created about one aspect of an individual to influence opinion on other areas in-group bias people favour members of the in group over the out group gambler's fallacy belief that past results influence future random events. hot hand fallacy mistaken belief that a person who has experienced some success has a greater chance of success at further attempts confirmation bias- tendency to seek out , interpret and create information that conforms to existing beliefs false uniqueness effect- tendency to view oneself and one's own qualities as unique Dunning Kruger effect people are inclined to believe themselves as more competent than they are. people with lowest ability tend to overestimate themselves the most. poor performers lack expertise to know how bad they're doing.

how to do chi square in JASP (detail) qualitative data

go in to Jasp and click on urban rural or whatever demographic item where you're changing the 0/1's in to words. click on the top of the row. change the label of the value eg 0 to rural and 1 to urban. do this for the other demographic item rows go in to frequencies and select contingency tables select the conditions asking about the use of masks (the last three) and put these in to columns if you select urban/rural and the other demographic items and put these as the rows This will give you the chi squared value for the demographic items and the three questions stacked on top of each other and the same for the next demographic item. Make sure cells> counts> expected is ticked. the significant values are <0.05. There is no significant difference between masks used in urban and rural areas for volcanic ash, but there is for urban rural for germ spreading to self and others. This may be because people in rural areas encounter less people. We know also that this isn't just because they're more readily available or that this isn't just a general trend in urban vs rural because it is the same for volcanic ash. men tend to wear masks less across all counts. Those who have had a respiratory disease where a mask more than those who have not. The P value is the chance that you would find that result is the null hypothesis was true. so if P= 0.017 there would be a 1.7% chance you would get that value if the null hypothesis were true so in that incidence you would reject null and accept the alternative. read the discussion section of the paper to see these results justified.

Iteration

is doing a process over and over again By immersing yourself in the data the line blurry between what you see in the data and what the data actually says. We want the data to speak for itself and not for us to randomly create meaning.

introduction

justify the hypothesis introduction acts as a big funnel, with the hypothesis at the end (the most narrow bit) so start broad and get more specific discuss general issues review relevant literature/ background general description of the experiment hypothesis be selective with what you include make sure your hypothesis is justified include earlier studies. you need an angle- why do you need to do this experiment? population, time, if its a replication -why

when do we use chi square?

make sure you know your research scenario and your data. Is this data quantitive or qualitative? Nominal and ordinal data are both qualitative. nominal data has different values that are things such as name labels in a non-specific order ordinal data is where the different values can be put in order but there is no regular interval quantitative is any data that can be counted and expressed manually. clearly identify your data and know its level of measurement. This will help you to know what stat test to use and what conclusions to make. If you do not properly identify your data then your analysis will be wrong. The chi squared can be used when there is nominal or ordinal data in one or two categories. The assumptions you make by using the chi square are the data is categorical, nominal, or ordinal variables consist of two categorical independent groups have to analyse data in numerical frequencies or counts Bivirate categorical data is summarised in a 2 way contingency table. ie you would have a row for nationality and different columns for different drink choices. Then add the total of the rows to make the observed joint frequencies. looking at frequencies alone can be misleading. This is because unequal group sizes make comparisons difficult. so you can ask yourself what the cell counts would have been expected to be if the two groups were the same size. Then compare this with the observed value. if there is a large difference between the observed and expected values then there is a difference between groups, if there is a small difference between the observed and the expected values there is no difference between the groups.

paired sample t test

matched pair/ correlated pairs etc- run t test on dependant samples. dependant samples are usually connected within subject design eg if you were using two different creams use one arm and use other. or same person at different points in time. when you're using a within subject design interval or ratio data no outliers approximately normally distributed. ratio between mean distance and sample variance is it one tailed or two tailed? what is our null and alternative hypothesis? null hypothesis is the population mean difference between the paired values is equal to 0. alternative the population mean difference between the paired values is not equal to 0. t value, degree of freedom, p value, cohen's d (effect size)

independent t test

most common form of t test helps to compare means of two sets of data between subjects ( separate groups for independent variable or independent measures design subjects are only in one group of experiment dependant variable must be interval or ratio no significant outlier outlier is observation that lies at an abnormal distance from other values in a random sample from a population can look at graphical representation and see eg look at box plot or line graph approximately normally distributed. determine if test is one tailed or two tailed determine your hypothesis H0 and H1 null- score will be equal no difference alternative- there will be a difference between the two groups if you don't know the direction or difference use 2 tailed, if you know use 1 tailed so we need the t value, degrees of freedom and p value. also need effect size or Cohen's D this measures the strength of the difference. small effect= 0.2 (difficult to see with naked eye when looking a graph) medium= 0.5 (can see with naked eye) large effect= 0.8 (can easily see)

what does data need to be to use a parametric or non parametric test

nominal and interval are qualitative - non-parametric tests interval and ratio are quantitative - parametric tests

why are there no degrees of freedom when calculated in JASP?

not included because they are not needed you should just do a central tendency measure and report sample size in your write up. rank biseral correlation is equivalent to Cohen's D. The general rule of thumb for biseral correlation is 0.1-0.3 is small effect size 0.3-0.5 is medium effect size 0.5-1 is a large effect size

one sample t test in Jasp

one column with all data go on to descriptives, look at the variable look at distribution plots and box plots look at outliers go in to t tests - one sample t tests still doing student t test put test value in (the known value that was predetermined before) say if one or two tailed tick effect size

do you do a one tailed or a two tailed test?

one tailed - a difference two tailed- difference but direction is specified- you make an assumption

how to use Jasp for independent samples t test

open data IV in one column, DV in one will tell you what kind of data it is but if this is incorrect you can change it by clicking scale, nominal or ordinal. go in to descriptives select dependant variable press arrow, to see table. always check data before doing statistical test. now to split the groups select the independent variable and press arrow in to the split window. press statistics drop box, can go on to shapiro test if you go on to plots drop box, can see distribution plots and display density. can also have a look in outliers by selecting box plots. and select label outliers. jitter element allows you to see bread of all values on box plot. now click on t test bar at the top. and then independent sample t test. and three different types of test. use student welch- adaptation when you have unequal samples./ homogeneity of variance is not met. mann-whitney- non-parametric version. alt hypothesis section - two tail select first and one tail either of the bottom 2. click effect size, click cohen's d. dependant variables in variables widow and IV in grouping variable. assumption checks- click normality. and equality of variance. can have a look in descriptive plots to visualise it.

Jasp experiment T test example

open data firstly look at the difference starting between heterosexual and non- heterosexual use an independant t test. you put time 1 (the first time they were assessed in to the variable window. you then go back to page 1 on Jasp and click on the groups and take away the control so you just have het and non het. Then put the group variable as 'groups' you can see from the results that there are no differences between them now we want to look at the changes in the scores or both groups before and after the program for this we select a paired t test if you put the variables as times one and two this gives you the results for the the whole data set as oppose to the separate groups so you need to go back to the first page, click on groups and then deselect one and get the results for the other and vice versa.

Spearman's Rho in JASP

open data line can sometimes go beyond P value as predictive select spearman's click display pairwise, report significance, flag significant correlations, can also add scatter plot and stats for reporting if you have a non linear scattergram there are special tests for non linear relationships covered at masters level

how to use Jasp for a paired sample t test

open data file. one column for each condition click plus and add a new coloumn name it label what type of data it will be allows arythmical operation then do difference click first minus second and then compute column go in to descriptives and then difference will be in variable window can look at distribution plot and box plot. go in to t test and paired sample t test put un to variable pairs the two you're looking at (not the difference) click wether it is one table or two tailed. and then output at bottom. can look at descriptive plot to visualise difference.

JASP and descriptive statistics

open up the CSV file open it on JASP and synch the data click on descriptives you will get two windows here 1 for the different groups and the variables and on the right hand side there will be the results . the < > arrows on the side of the page allow you to flick back to the data table and back to the descriptive stats. You can also use the three dots on top of each other to view all three windows at once. Click on group one and then click the small triangle in the middle of the groups and the other box labelled variables. Then when you look over to the results it will show you the mean, standard deviation etc of that group. if you want to see other descriptive stats like inter quartile range there is a box that says statistics. click on this and it will drop down your other options. If you want to then see the data for groups 2 and three press these and the arrow again. This will let you see all three sets of data next to each other in different coloumns. to highlight all of the groups at once press command then click. You can remove groups data by selecting the group and pressing the triangle again which will now be pointing in the other direction. You can create three different tables for the different groups by pressing the green + button for the number of groups you want and removing the irrelevant groups from being shows in that table. Press then black pen icon to rename the tables. This allows you to make three separate table sin a vertical column. If the file is synched correctly and you change a value, JASP should not only change it in the table but also in the descriptive stats too.

what is a t-test?

parametric test use to compare mean of two groups how significant is the difference and if this difference should happen by chance also known as student's t test data must be ratio or interval compare 2 groups makes a t score which is the difference between the two groups and the difference within the groups if t value is 0 then the data is the same, larger the t score, the greater the difference

intro to non-parametric tests

parametric tests use interval or ratio data, and data that is also normally distributed and has homogeneity of variance You cannot use a parametric test if these conditions are not satisfied (well you can but it isn't that good) non-parametric tests have no assumptions about the data, this means you can use a non-parametric test on data that's not normally distributed. If the data is categorical it can't be normally distributed because the interval between each value is not necessarily the same. ordinal data is a kind of categorical data with a set order to it eg a likert scale. but the gaps between the values are not meaningful. ordinal is not standardised - you need to look at the mode, median, interquartile range and bar charts but NOT histograms and the mean. We also use non-parametric tests on small sample sizes or when data is interval or ratio but not normally distributed or the data is categorial. rank data by putting in to numerical order and then find the lowest score and give it a rank of one then go up etc. The analysis is carried out on the ranks rather than the data itself. so is non-parametric less powerful? ranking data allows a way around distributional assumptions of parametric tests but but ranking data looses information about the magnitude of differences between scores. statistical power is the ability of a test to find an effect that genuinely effects the data, as oppose to non-parametric but only when the normal data assumption is met. 4 common types of non-parametric test are 1. mann whitney U 2. Wilcoxon signed rank. 3. Kruskal-Wallace 4. Friedman

methods

participants, apparatus, design, procedure participants- age, gender, relevant info, any sight problems?- corrected to normal vision? apparatus- questions used or asked , if used include technical equipment, list of words, stimuli, instruments, questionnaires IV's and DV's should be included in the design procedure- this should be a complete account of what you have done. How participants were recruited for experiment, how they were told about it, the participant instructions can go in the appendix if you say see here.

How to do JASP 1 sample test

put difference in separate column calculate N go in to distributions, select binomial distribution probability of success is 50/50 if 2 outcomes so will be 0.5. number of trials is n value (without zeros) click calculate probability

Discussion

restate the hypothesis and say if you found evidence to support it. start narrow and end broad discuss the results- do the results support the conclusion? does previous research support the conclusion? what does this mean? how can the results be interpreted in line with current theory? are there multiple interpretations? what other studies might need to be done? what are the problems with the theory? always assume your data is right. -relate to hypothesis -relate to previous research -relate to theoretical models -problems DO WIDER READING AROUND THE TOPIC you don't need numbers or to repeat the results. don't describe the data in detail. support and criticisms if any odd or unexpected results discuss how or why this may have occurred. is it important? why are the results interesting? limitations bring in more relevant research (past 10 years) references APA format show evidence of wider reading need recent studies from the past 10 years

t test distribution

sampling distribution. plot them without having to collect may sample. degrees of freedom is closely related to sample size. different t distribution for every sample size, option for this in JASP. why is it important? place calculated value in t distribution to see how consistent results are with no hypothesis and will tell you wether to direct or accept. each t value has a p value the p value is the likelihood that your results occurred by chance. low p values are good. accepted p value is 0.05 if it is above we conclude results are not significant.

An introduction to ethics in research

sometimes it is important to talk about the experimenter/ researcher in an experiment. eg the stanford prison experiment (1971) This is where Zimbardo investigated deindividualisation, an anti-social behaviour. This was valuable research and could explain atrocities such as Abu Ghraib (2003) and Milgram (1963). So the experiment gave us knowledge that we would not have obtained without it. Zimbardo himself had a strong role in the stanford prison experiment. Recordings show Zimbardo's assistant coaching guards to be 'tough' on prisoners. This may have lead to bias in the experiment. This experiment illustrates an example of a study where from the get go it was likely for people to come to harm. There was in fact an ethics committee that approved the experiment, concluding that the gains from the research would be greater than the harm done to individuals. Things that made the experiment unethical in particular were the: care of participants- subjected to psychological and physical torment quality of reporting- Zimbardo omitted details about coaching guards bias in research- Zimbardo thought he was justified in his actions People, even researchers are political. So there will always be some inherent political bias in research. There is a balance between the gains made from the experiment and the harm done to individual participants and society. Research ethics- The BPS requires that we must have some general understanding of ethics in society. All experiments must be approved by an ethics committee.

Measrures of spread- standard deviation, interquartile range, and overall range

standard deviation is the measure of variability in scores of a variable. It is used to give us an indication of how scores vary around the mean score. Essentially the average deviation from the mean. This is the equation for the standard deviation where X is an individual score, X with a line on top is the mean score, n is the number of data points. sigma is the sum of. There is also the computational formula which is the easier formula. The first step is to calculate the mean score. First you do the bit in brackets so you get the first individual score in the data set and then subtract the mean from it. You then square this, you do this for all the other values in the data set. You then add these all together. You square the values to avoid any negatives as these would cancel out the positives making the data meaningless. n-1 is the degree of freedom this is the number of values in the data set -1 and so you divide the sum of all the values from the mini equation above (the sum of squares) by the degrees of freedom. You then find the square root of this value. We do this so that we essentially undo the effects of the squaring in the beginning. This actually makes the data more interpretable thus we now have the average deviation from the mean score The inter-quartile range is used when the median is used as an average instead of the mean this is used when there are extreme values or the data is very skewed. The first step is to calculate the median. The interquartile range is the spread of the middle half of the data. the lower quartile is the median scores to the left of the median when ordered from low to high. The upper quartile is the median scores to the right of the median score, when ordered from low to high. Must calculate the value at the 25% mark and the value at the 75% mark. The interquartile range is the upper quartile number minus the lower quartile number. standard deviation vs interquartile range- standard deviations are most commonly used when using parametric tests which are considered the most powerful kind of test. Additionally the standard deviation is the more precise estimate of variation. it takes in to account all data from the set. However, as with the mean it is heavily influenced by extreme scores and skewed distributions. The interquartile range excludes extreme scores, and is usually a good representation of how visual data is represented eg box plots. However, it uses arbitrary cut off points, and half of the data is basically ignored. the overall range is the highest value in the data set take away the lowest value in the data set. it is rarely used in statistical analysis but it is good to check in JASP so you can get an overall idea of all the range of scores and also to check you haven't made an error eg avoid typos in data entry.

results

the numbers, no interpretation sentence at the brining briefly about what there is eg here are the generalised mean participant scores and the subsequent paired T test. Descriptive statistics should come first and then inferential statistics descriptive stats should be in a table or graphical display tables and graphs should be correctly formulated all raw data goes in the appendices not the results. only describe the results, interpretation is left for the discussion section. sentence or two at the start table 1 shows the mean scores and standard deviations This is then shown in a graph in figure 2. interpretative statistics eg wilcoxon's

implicit association test

this is a reaction time test with the basic idea that the quicker you sort things the more strong the association is. 80% of people have a bias for young people vs older people 69% of people have a bias for thin vs overweight 68% have a preference for white people over black people 76% of people have a preference for able bodied people over disabled people implicit stereotypes 72% of people stereotype black people as being associated with weapons 76% stereotype women as caregivers criticisms of this test ambiguous if it is state or trait designation- are the results malleable or not? can be strongly affected by the environemnt- low test-retest reliability - but is this what bias actually is? is this measuring bias or cultural awareness? individuals can control implicit responses? Wallerant et al no evidence that the results form these tests alter behaviour

How to do Pearson's r in JASP

to check if there is a linear relationship we need to look at the scatter graph go in to descriptives, plots, correlation plots, change order of variables to change which axis the values are on then go to regression, correlation click pearson's 'r' click on your selected hypothesis click report significance you can also click display pairwise and click flag significant correlations if you have a lot of data (this shows you what you're interested in) click statistics and add P value if you want to display something to use in a report

how to use a parametric test

to use parametric tests look at different assumptions scale of measurement must be interval or ratio data approximately normally distributed how to make sure data is normally distributed sample size ( if number is bigger than 30 chances are that it is normally distributed) look at histogram distribution of data shapiro-wilk's test - tells you if data is normally distributed or not. if test is not significant p> 0.5 then normality is assumed. homogeneity of variance can look at with naked eye but there is also a test for this. ( will learn in second year) so once you know what type of data is, you need to know what kind of question you're asking. relationship- aim to study the causal relationship between variables. 2. differences- aim to compare two or more groups ( is there a difference) so now- what are your variables? dependant variable is the outcome of your data. eg the thing you aim t affect though your change. ( what you measure) independent variable- factor you manipulate to affect the dependant variable now how many levels do you have in your independent variable? difference- is there a difference between these two groups? two tests or multiple tests changes which tests you use

how to report pearson's r

when you report you need to say what test you did, what the two variables were and the 'r' and p values

how to write up spearman's rho

write up - stats, correlation coefficient, P value, describe strength an direction of relationship.

examples of qualitative analysis

•Content analysis •Thematic analysis •Discourse analysis •Conversation analysis •Inductive thematic analysis •Grounded theory

Interrater reliability

•How do we know I am using codes in the same way as you? •One answer: Percentage Agreement (do I select the same codes as you?) •One problem: What if codes are very common or very uncommon? if very common would just agree a lot by canvce because you would find them more, whereas if very uncommon would be many zeros so would thus agree more. Agreement could be high by chance. •Another answer: Cohen's kappa - a test statistic that takes into account chance agreement (ranges up to 1; 0.6 = moderate, 0.8 = good, use for 2 raters) Other stats: weighted kappa, Fleiss' kappa, Krippendorff's alpha (for 2+ raters inter-rater reliability is establishing the principle that when one person codes, coding should be consistent and transparent so that another person should be able to code in the same way. Cohen's kappa- given how frequently a code is used what's our expected provability, and given low frequently how many times it has been agreed. This must be above the expected value. not all qualitative researchers think you should do this, in psychological journals it is expected so people know your results are reliable. if not doing anything quantitative wouldn't necessarily be expected.