Statistics Micro Mini Threats to Your Experiment!

Slides:



Advertisements
Similar presentations
Experimental Design.
Advertisements

Agenda Group Hypotheses Validity of Inferences from Research Inferences and Errors Types of Validity Threats to Validity.
1 COMM 301: Empirical Research in Communication Lecture 15 – Hypothesis Testing Kwan M Lee.
MAT 1000 Mathematics in Today's World. Last Time 1.What does a sample tell us about the population? 2.Practical problems in sample surveys.
Validity of Quantitative Research Conclusions. Internal Validity External Validity Issues of Cause and Effect Issues of Generalizability Validity of Quantitative.
Decision Errors and Power
Statistical Decision Making
Copyright © 2010, 2007, 2004 Pearson Education, Inc. Chapter 13 Experiments and Observational Studies.
Correlation AND EXPERIMENTAL DESIGN
Slides to accompany Weathington, Cunningham & Pittenger (2010), Chapter 4: An Overview of Empirical Methods 1.
Empirical Analysis Doing and interpreting empirical work.
Validity, Sampling & Experimental Control Psych 231: Research Methods in Psychology.
Validity, Sampling & Experimental Control Psych 231: Research Methods in Psychology.
Non-Experimental designs: Developmental designs & Small-N designs
MSc Applied Psychology PYM403 Research Methods Validity and Reliability in Research.
Non-Experimental designs: Developmental designs & Small-N designs
Math Candel Maastricht University. 1.Internal validity Do your conclusions reflect the “true state of nature” ? 2.External validity or generalizability.
Sampling and Experimental Control Goals of clinical research is to make generalizations beyond the individual studied to others with similar conditions.
PSY 1950 Confidence and Power December, Requisite Quote “The picturing of data allows us to be sensitive not only to the multiple hypotheses that.
Psych 231: Research Methods in Psychology
C82MCP Diploma Statistics School of Psychology University of Nottingham 1 Overview of Lecture Independent and Dependent Variables Between and Within Designs.
Using Statistics in Research Psych 231: Research Methods in Psychology.
The t Tests Independent Samples.
Impact Evaluation Session VII Sampling and Power Jishnu Das November 2006.
EVAL 6970: Experimental and Quasi- Experimental Designs Dr. Chris L. S. Coryn Dr. Anne Cullen Spring 2012.
Experiments and Observational Studies.  A study at a high school in California compared academic performance of music students with that of non-music.
Chapter 8 Experimental Research
Experimental Design The Gold Standard?.
ANCOVA Lecture 9 Andrew Ainsworth. What is ANCOVA?
Chapter 13: Inference in Regression
Chapter 2: The Research Enterprise in Psychology
Chapter 8 Hypothesis testing 1. ▪Along with estimation, hypothesis testing is one of the major fields of statistical inference ▪In estimation, we: –don’t.
Chapter 8 Introduction to Hypothesis Testing
Chapter 1: Introduction to Statistics
Day 6: Non-Experimental & Experimental Design
Copyright © 2010 Pearson Education, Inc. Chapter 13 Experiments and Observational Studies.
Experiments and Observational Studies. Observational Studies In an observational study, researchers don’t assign choices; they simply observe them. look.
Chapter 13 Notes Observational Studies and Experimental Design
Chapter 13 Observational Studies & Experimental Design.
Copyright © 2007 Pearson Education, Inc. Publishing as Pearson Addison-Wesley Chapter 13 Experiments and Observational Studies.
+ Chapter 9 Summary. + Section 9.1 Significance Tests: The Basics After this section, you should be able to… STATE correct hypotheses for a significance.
Chapter 8 Introduction to Hypothesis Testing
Chapter 2 The Research Enterprise in Psychology. Table of Contents The Scientific Approach: A Search for Laws Basic assumption: events are governed by.
Chapter 5: Producing Data “An approximate answer to the right question is worth a good deal more than the exact answer to an approximate question.’ John.
Slides to accompany Weathington, Cunningham & Pittenger (2010), Chapter 3: The Foundations of Research 1.
Independent vs Dependent Variables PRESUMED CAUSE REFERRED TO AS INDEPENDENT VARIABLE (SMOKING). PRESUMED EFFECT IS DEPENDENT VARIABLE (LUNG CANCER). SEEK.
1 Chapter 10: Introduction to Inference. 2 Inference Inference is the statistical process by which we use information collected from a sample to infer.
Chapter 3.1.  Observational Study: involves passive data collection (observe, record or measure but don’t interfere)  Experiment: ~Involves active data.
Experiment Basics: Variables Psych 231: Research Methods in Psychology.
Chapter 6 Research Validity. Research Validity: Truthfulness of inferences made from a research study.
Experimental Research Methods in Language Learning Chapter 5 Validity in Experimental Research.
The Practice of Statistics, 5th Edition Starnes, Tabor, Yates, Moore Bedford Freeman Worth Publishers CHAPTER 4 Designing Studies 4.2Experiments.
Experimental & Quasi-Experimental Designs Dr. Guerette.
Research Design. Time of Data Collection Longitudinal Longitudinal –Panel study –Trend study –Cohort study Cross-sectional Cross-sectional.
Introduction to Validity True Experiment – searching for causality What effect does the I.V. have on the D.V. Correlation Design – searching for an association.
Producing Data: Experiments BPS - 5th Ed. Chapter 9 1.
1 Chapter 11 Understanding Randomness. 2 Why Random? What is it about chance outcomes being random that makes random selection seem fair? Two things:
Some Terminology experiment vs. correlational study IV vs. DV descriptive vs. inferential statistics sample vs. population statistic vs. parameter H 0.
Chapter 5 Data Production
CHAPTER 4 Designing Studies
Section Testing a Proportion
CHAPTER 4 Designing Studies
Chapter 6 Research Validity.
CHAPTER 4 Designing Studies
CHAPTER 4 Designing Studies
CHAPTER 4 Designing Studies
CHAPTER 4 Designing Studies
CHAPTER 4 Designing Studies
CHAPTER 4 Designing Studies
CHAPTER 4 Designing Studies
Presentation transcript:

Statistics Micro Mini Threats to Your Experiment! January 7-11, 2008 Beth Ayers Threats to Your Experiment

Threats to Your Experiment Here I am using threats to mean things that will reduce the impact, credibility, or generalizability of the results your experiment or study, particularly the things that we have control over Threats to Your Experiment

Threats to Your Experiment Validity Statistical conclusion validity Internal validity Construct validity External validity etc. Type I Error Power Threats to Your Experiment

A note of caution Many of these ideas are inter-related I expect some of this to be confusing at first, it will help to reread the notes once you’ve seen all the terminology Threats to Your Experiment

What is Validity? In general, validity refers to the degree to which a study supports the intended conclusion drawn from the results Validity is broken down into many different types A Google search will yield many pages explaining different types At a quick glance Wikipedia articles seem to have many of the types and are correct Threats to Your Experiment

(Statistical) Conclusion Validity Conclusion validity is the degree to which conclusions we reach about our data are reasonable In particular, is there a relationship between the two variables? Threats to Your Experiment

Threats to Conclusion Validity Concluding there is a relationship when in fact there is not one – Type I Error Concluding there is not a relationship when in fact there is one – Type II Error Low statistical power Violation of assumptions Fishing for results Threats to Your Experiment

Controlling Conclusion Validity Threats When testing multiple hypotheses, adjust the error rate, see slide 20 Take reliable measurements Increase power (see later definitions) Use the right statistical test and perform it correctly!!! Threats to Your Experiment

Association does not imply causation! Internal Validity Internal validity is the degree to which we can conclude that the changes in the explanatory variable caused the changes in the response variable Remember Association does not imply causation! Threats to Your Experiment

Threats to Internal Validity Confounding – in the form of a lurking variable Selection Bias – before treatment there are differences between the groups History – some historical event affects the study outcome Maturation – if making observations at different times points, the natural aging of the subjects may invalidate results Instrument change – not using the same measurement tools or methods with each subject Repeated testing – subjects may remember answers between testing sessions Experimenter bias – the experimenter treats subjects differently based on treatment type Threats to Your Experiment

Controlling Internal Validity Threats Random assignment of treatment or randomization Blinding – subjects don’t know which treatment they are receiving Double blinding – neither subjects nor experimenters know which treatment subjects are receiving Threats to Your Experiment

Construct Validity Construct validity refers to the degree to which inferences can be made from the measurements in your study In particular, are you manipulating what you claim you are manipulating (the causal construct) and are you measuring what you claim you are measuring (the effect construct) Often defined with multiple subcategories of validity Threats to Your Experiment

Threats to Construct Validity Poor study design Using new or unreliable methods of measurement Measurements depend on who is collecting them Interaction of testing and treatment Interaction of treatments Threats to Your Experiment

Controlling Construct Validity Threats Carefully design your study Have others critique your design Measurements should be reliable and not depend on who is administering your test If there is a “gold standard” of measurement for your response variable, your measure should have a high correlation with that test Threats to Your Experiment

External Validity External validity is the extent to which we can generalize the findings to particular target persons, places, or times and to which we can generalize across types of persons, places, or times Often simply called generalizability Threats to Your Experiment

Threats to External Validity The particular group of people used in your study The particular place where you performed your study The particular time at which you performed your study Threats to Your Experiment

Controlling External Validity Threats Make sure to get a representative sample of the population Do your study in a variety of places, with different people, and at different times Threats to Your Experiment

Validity Questions Are Cumulative Conclusion – is there a relationship between the explanatory and response variables? Internal – is the relationship causal? Construct – can we generalize to the construct? External – can we generalize to other persons, places, times? Threats to Your Experiment

Type I Error Although this was mentioned in conclusion validity, it is worth discussing again Due to the inherent uncertainties of nature, we can never make definite claims from our experiments Retain H0 Reject H0 H0 is true Correct Type I error False discovery H1 is true Type II error Threats to Your Experiment

Type I Error However, we can set a limit on now often we will make a false claim Done by setting the ®-level of a test Do not “data snoop” Performing many tests on your data until you find a significant result Correct for multiple testing Threats to Your Experiment

Correcting for Multiple Testing When performing hypothesis tests, 100*® will result in Type I errors When doing many tests, need to lower the error rate Several common methods are: Bonferroni correction Benjamini-Hochberg method False discovery rate Threats to Your Experiment

Statistical Power The power of an experiment refers to the probability that we will correctly conclude that the treatment caused change in the outcome given that it actually does Low power leads to Type II Errors Have some control of the power of your experiments prior to running them Performing experiments with low power is a waste of time and money! Threats to Your Experiment

Statistical Power The power varies from ® to 1 Typically people agree that 80% power is a minimal value for good power Poor power is a common problem. It can NOT be fixed with statistical analysis. It must be dealt with before running your experiment. Threats to Your Experiment

Statistical Power Statistical power depends on the ®-level of the test the size of the difference or the strength of the similarity (that is, the effect size) in the population the sensitivity or variability of the data More power to detect larger effects than smaller ones Threats to Your Experiment

Ways to Increase Power Increase sample size Reduce variability Increase the spacing between population means Threats to Your Experiment

Increasing Sample Size Unfortunately, this is not always an option Not always cost effective Threats to Your Experiment

Reducing Variability Types of variability Measurement Environment Treatment application Subject-to-subject There is a trade-off between reduced variability and external validity If we to do much to control the variability, we lose generalizability Threats to Your Experiment

What can power calculations do? Suppose we’re creating an on-line tutor to help students study for the SATs. Since we’d like to sell the package for $200, we believe that we’ll need a 100 point increase in SAT scores in order for parents to consider the tutor worthwhile. Given a sample size, an ®-level, and an estimate of the variability of SAT scores, we can calculate the power of our experiment Given a power we’d like to have, an ®-level, and an estimate of the variability of SAT scores, we can calculate the sample size needed to obtain that power Threats to Your Experiment

Notes on Power Calculating power can be hard… However, there are many on-line applets to help you calculate power! As we go through examples I will show you one such applet Find one you understand and can use! Threats to Your Experiment