Research Methods 1 Lecture 1 2023-24 PDF

Research Methods 1 Lecture 1 – Introduction to Research Design and Data Module Overview - Lectures Lectures Introduction to Research Design and Data Significance Testing Independent Samples T-Tests Paired Samples T-Tests Tests for Nominal Data Correlation...

Research Methods 1 Lecture 1 – Introduction to Research Design and Data Module Overview - Lectures Lectures Introduction to Research Design and Data Significance Testing Independent Samples T-Tests Paired Samples T-Tests Tests for Nominal Data Correlation Qualitative Methods Module Overview – Skills Practicals PC lab sessions to help you practice what we learn in lectures Using Excel and Descriptive Statistics Creating Figures and reporting results Intro to SPSS Independent Samples Tests Paired Samples Tests Nominal Tests Correlation Which Test? – Choosing Appropriate Analyses Rinse & Repeat! Assessments MCQ Exam Questions about all the content from the course You will need to interpret data outputs provided to you You won’t need to do any statistical analysis live in the exam! We are interested in testing your understanding of the concepts and methods It’s not a maths test VLE tests Open book online tests Released on the VLE and submitted (usually) a week later You can work on it at your own pace You will need to conduct analyses during these tests, but it is not timed VLE test deadlines See the Wiki for deadlines and release dates You will need to submit your test before noon on the deadline date You should use the discussion board for questions, but I won’t answer any questions directly related to the test during the open test window Reading “You don’t have to know everything, but you should learn how and where to find the things you need and want to know.” -Sophonisba Breckinridge Software We will use Excel to organise our data and create graphs We will be also be using a software package called SPSS for all our analysis needs Obtaining Software You can download the software you need from the university website What if I don’t like Maths? Research Methods is more than just maths! Planning Experiments Thinking carefully Interpreting data Questioning and thinking critically Creativity Finding new ways to test and answer questions Research methods is a tool Why is Research Methods Useful? Science is all about asking and answering questions Research methods helps us: Ask good questions Design experiments and gather data Interpret the data to answer your questions Communicate ideas effectively Teaching you to become independent researchers My Pledge If you engage with the lectures and practicals there won’t be any surprises - You might even find you learn to love it! I am not trying to trick you – the exam will test you on what we have learnt together this semester. I am available to help ○ Please let me know if you are stuck or struggling with anything ○ Put your hand up in a practical and I will come over and help one-to-one ○ You can also contact me via email or the discussion forum – don’t be shy Trust me, and reach out if you need to This Session Research Design ○ Variables ○ Experimental Design Quantitative Data ○ Data Types ○ Descriptive Statistics Qualitative and Quantitative Research Quantitative Research is all about numbers Qualitative Research goes beyond the numbers ○ Interviews, Open Ended Questionnaires, Focus Groups etc Both methods are important and often a combination approach is used This course will teach you about both types of research ○ My lectures focus on quantitative methods ○ Maurice Waddle will teach you about Qualitative methods later this year Variable Types Independent variable: The one you change / manipulate Dependent Variable: The one you measure Control Variables: Things you intentionally keep the same in your experiment Control Condition: A separate condition that helps you understand the role of the IV, or help you rule out alternative explanations for your results. Experimental Design Within-Subject Design: ○ Each participant takes part in all conditions ○ → Each participant contributes multiple data points ○ Also called Repeated Measures design ○ Accounts for individual differences ○ Often cost and time effective (can recruit half as many participants for same number of datapoints in each group) ○ Can introduce issues with order effects/ fatigue effects etc Experimental Design Between-Subject Design: ○ Different participants in all conditions ○→ Each participant contributes one data point ○ Avoids some participant/experimenter effects ○ Order effects ○ Fatigue effects ○ Etc ○ Takes longer and is less powerful than a within-subject design ○ Introduces variation due to individual differences Experimental Design Matched Pairs Design: ○ Different participants in all conditions ○ Participants matched across conditions to try to account for individual differences ○ E.g. Gender, Experience, Score on previous task etc. ○ Has a lot of the benefits of both methods ○ but it is often difficult to match people accurately. Research Design Example You want to understand how the time of a class influences students’ memory for the material. You decide to compare the exam results for 3 core modules that have classes at different times. The same students take all three exams. What is the independent variable? Time of class What is the dependent variable? Exam Results What type of design is this experiment? Within-subject design Experiment Example Class 1: Research Methods at 9am on Thursdays Class 2: Psychology in the Movies at 1pm on Monday Class 3: Criminal and Forensic Psychology at 4pm on a Friday Conclusion: Later classes lead to better exam results than earlier classes Confounds and Extraneous Variables The example experiment has many variables that were not controlled: Class Type: Research Methods, Movies, Forensic… ○ It is likely that on average preferred Movies and Forensic over RM Different Lecturers ○ Teaching styles ○ Personality ○ Attractiveness!? ○ Incentives - gives out chocolates/puts funny gifs in slides… Different days of the week ○ Mondays are hard ○ Lectures the morning after student night are more poorly attended Confound Vs Extraneous Variables Extraneous Variables: ○ Not controlled in the experiment ○ Could have an effect on the DV ○ e.g. Weather or Noise outside testing room Confounding Variables: ○ Extraneous variables that vary systematically with the IV to influence the DV. ○ Confounding variables are likely to influence the results ○ e.g. One condition much longer and more boring than the other → Fatigue effect → Variance in the DV can appear to be due IV when it is not What could really be happening in our study? Time of day influences results? Subject preference? Day of the week - hangover? RM Film Forensic There are too many alternative explanations for this result to trust the conclusion that later classes are better... Design Summary Picking a design is an art! Design depends on your question and your variables Within vs Between - consider the benefits and drawbacks of each Controlling extraneous variables is important to avoid confounding the results When discussing confounds - ask yourself if the variable could really have affected the results or not. You will need to practice thinking about experimental design Data Types Data Types are important For SPSS... (more on this next week) For checking your data meet the requirements for certain statistical tests. E.g. You can’t run a correlation analysis on nominal data. Data Types: Categorical Categorical data → Labels Sometimes called Discrete Data Nominal (Name) Category Labels with no hierarchical order e.g. Gender, Nationality, University etc Ordinal Categories with a hierarchical order E.g. First Second Third place, Year Group, Illness Stage etc The distance between the points on the scale are not Data Types: Continuous Continuous data is data on a scale Interval Scalar/Continuous data that has no meaningful zero The intervals on the scale are equidistant E.g. Temperature in Fahrenheit Ratio Scalar Continuous data with an absolute zero The intervals on the scale are equidistant E.g. Time Descriptive Statistics Descriptive Statistics are numbers that summarise our data ○ Mean, Median, Mode, Standard Deviation etc You can present Descriptive Statistics in text, tables or graphs It gives the reader an idea of the pattern of your data, without giving all the raw scores You should present your data according to APA formatting guidelines Descriptive Statistics It is worth revising how to calculate these measures manually or in Excel Mean  Add all data points up and divide by the number of data points In Excel =average() 21, 34, 31, 28, 24, 26 Total = 164, Count = 6 164 / 6 = 27.33 Descriptive Statistics It is worth revising how to calculate these measures manually or in Excel Median  Middle point of an ordered list In Excel = median() 21, 24, 26, 28, 31, 34 26+28 = 54 54/2 = 27 Other Descriptive Statistics Standard Deviation = stdev() Standard Error = standard deviation / square root of N Range = Maximum – Minimum You will need to be familiar with calculating these numbers For the VLE test For the Exam – you will have a calculator Z-Scores A Z score is a standardized score It represents a datapoint’s relationship to the mean of a group of values It is useful for comparing scores between participants Or across conditions Useful particularly when units differ E.g comparing a score out of 10 vs out of 25 Or comparing reaction time scores with accuracy scores etc Z = (score-mean)/standard deviation Z-Score Example My Class Class Z- Subject Score Average SD Score Maths 65 68 8 -0.375 Music 68 71 6 -0.5 0.8333 Chemistry 54 44 12 33 Even though Music was my best raw mark, compared to my peers I did the worst in music and the best in Chemistry Which descriptive statistics should I use? The type of statistics you report will depend on the data type Nominal Data can be presented as frequencies ○ In our sample 10/12 participants responded that they preferred branded cola compared to 2/12 who preferred unbranded cola. Which descriptive statistics should I use? Continuous data should usually be presented as means and standard deviation either in text, in a graph, or in a table (pick one). ○ On average, participants had faster reaction times after drinking coffee (Mean = 1.12 seconds, SD = 0.30) than after drinking herbal tea (Mean = 2.34 seconds, SD = 0.51). Which descriptive statistics should I use? Continuous data should usually be presented as means and standard deviation either in text, in a graph, or in a table (pick one). 2.5 2 Reaction Time (Seconds) 1.5 1 0.5 0 Coffee Herbal Tea Drink Type Figure 1. Mean reaction time (seconds) after drinking coffee (left) and herbal tea (right). Error bars represent 1 Standard Deviation Which descriptive statistics should I use? Continuous data should usually be presented as means and standard deviation either in text, in a graph, or in a table (pick one). Table 1. Mean reaction times after drinking Coffee and Herbal Tea. Numbers in brackets represent Standard Deviation. Coffee Herbal Tea Mean 1.12 2.34 (SD) (0.30) (0.51) Distribution of Data If our data is normally distributed it falls into this shape known as a Bell- Shaped Curve This bell curve has the majority of data in the middle around the average point There is some data above, and some data below and the tails taper off evenly on both sides Normal Distribution of Data Some statistical tests require our data to be normally distributed We will learn how to test for normality using a test called Shapiro-Wilk (later this semester) Normality is important as it gives helps us predict the population mean from a sample mean We use a sample of data in our studies, but really we are interested in understanding the whole population! Confidence Intervals Confidence intervals are often used as a measure of the spread of data Instead of range or standard deviation etc A confidence interval has an upper and a lower number We are 95% confident that the population mean lies within this range 1.96 standard deviations either side of the mean represent the 95% CIs 1.96 is a z-score! Confidence Intervals = Mean ± (1.96*Standard Error) Non-Normal Distributions Sometimes, our data is not normally distributed Analysing Non-Normal Data If your data is not normally distributed, the mean is unlikely to represent the data accurately This means that statistical techniques that use the mean, may lead you to the wrong conclusion If your data is skewed, the median better represents the average The first choice of statistical test usually works on the means (parametric tests) If your data fail the assumption of normality, we can use alternative tests that look at the medians instead (non-parametric tests) Summary Research Methods won’t be too scary and I’m here to help Choosing a design for your question is an art and a science! Design type (within/between subject design) Data type (Nominal/ Continuous) Variables (Independent/Dependent/Control/Confounds) Descriptive Statistics summarise our data Means, Standard Deviations, Medians, Range, Z-scores Normal and non-normal Distributions Bell Shaped Curve, Skewedness, Confidence Intervals