1 17 15 Experiments I Ethical dilemmas a Experiments are rare in criminology criminal justice because you would have to experiment on people i ii If your experiment doesn t work then you end up intervening in a harmful way If the the treatment does work the control group is receiving the beneficial treatment II The purpose of an experiment is to guarantee time order and no alternative explanations a To run the experiment lab external validity problems arise III True Experiments a The experimenter in charge of the cause I V i Manipulates the independent variable I V and then observes the dependent variable thus cause precedes the effect ii Tries to ensure that nothing varies besides I V no alternative explanation b If 1 And 2 Work and I V and D V are correlated we have a causal relationship i We made the independent variable happen we have the right time order we guarantee no influence from outside sources and we have computed correlation c Manipulating the I V fixes the time order i Controlling alternative explanations preserves internal validity d Correlation is either found or is not i We cannot make the correlation happen IV The classical experiment a I V and D V i Not all causes can be manipulated so sometimes true experiments are not possible b Pretest and posttest i Measure the dependent variable before pretest and after posttest manipulating the I V ii Differences between the measurements are attributed to the influence of the I V 1 Measure of correlation c Experimental group and control group i Experimental group gets the I V ii Control group does not get the I V iii Control group measures how much the experimental group would have changed without the I V d A good Control group experimental group comparison requires groups that were identical to begin with i In a true experiment including classical experiment subjects are randomly assigned to the experimental group control group 1 This lets the experimenter say that all else are equal between the groups V Example of a Field Quasi Experiment What Would You Do TV Show a The subjects were the bystanders b What were the two groups i The group who interacted with the older actress ii The group who interacted with the younger actress d What elements of the classical experiment were missing c What was the I V The D V i D V helping the actors ii I V age of actors i No random assignment ii No pretest e Alternative Explanations i Diabetes vs infection ii Unemployment vs social security VI Control Groups Change Indicates Whether These 4 are issues a These would affect both the experimental group and the control group i History external events coincide with the independent variable affect the DV ii Maturation natural often age linked change in DV over time iii Testing taking the pretest may change the way subjects respond to iv the posttest Instrumentation measure of D V changes from pretest to the posttest 2 19 15 Quasi Experiments I True experiments feature control of alternative explanations through random assignments a Trade off between external and internal validity II Quasi experiments I V treatments is manipulated by you nature policy etc a b Cases not randomly assigned to treatment and comparison groups or conditions c Two types i Internal validity threats i Nonequivalent group design 1 Probably comparing group that are equivalent to begin with If they start off different it wouldn t be a surprise for them to end differently a 2 Make comparison group as similar as possible a Match individual cases on important variables b Create groups with comparable average characteristics i No longer have to have the exact individual that matches just need the average ii Ex on average these people are 21 5 years old and 51 female they have a certain racial breakdown 1 Just make a group that is equal on average 3 Often someone else a judge warden etc creates the groups a Determine who is allowed into programs generally not random 4 Threats to internal validity a Selection bias i Preexisting differences between groups or different potentials for change ii Could be an alternative explanation b Mortality attrition i Cases that drop out of the study ii Could happen in either quasi experiment or experiment c Statistical regression i Imperfect reliability sometimes a subjects pretest score will be randomly high or low 1 At posttest subjects score will tend to move closer to the average score regression to the mean 2 Problem when pretest scores guide subject selection assignment a Need to guarantee that the pretest is reliable ii Time series design 1 Examine a series of observations over time 2 Interrupted observations compared to before and after I V a Ex policy enacted treatment program start 3 Increases chances of instrumentation threat more time for changes in measurement to occur 4 Threats to validity a History b Instrumentation 2 26 15 Chapter 6 I Review of commonly problems from the quiz a Difference between experimental and quasi experimental i Experimental 1 Random assignment of treatment 2 Advantage a Cuts out sampling bias b The control and the treatment groups will be equal because of random assignment 3 Disadvantage a Because you are controlling the environment so much you do not have great generalizability ii Quasi experimental 1 No random assignment of treatment 2 Disadvantage a Internal validity issues the groups experimental and control may not be equal when the experiment is first started b Difference between comparison groups and pretest i Comparison 1 None equivalent a You are comparing the treatment group to the comparison but the groups are not necessarily an equal comparison b Use comparison group to describe the control group equivalent treatment group 2 Term specific to quasi experiments ii Pretest 1 Used to measure the dependent variable before the treatment is actually executed 2 The measure of the dependent variable is repeated again at the end of the experiment and you then compare the two measurements 3 Essentially the measurement of the effect of the dependent variable II Population III Master list a The the entire group of interest a List of all people in the population b Very hard to capturing every person in a large population i Ex the US Census is performed 10 years and every time there s is a large margin of error because millions of people are missed in the count IV Sampling unit V Why do we sample a The individual unit within the population b Depending on the population that is going to characterize the unit
View Full Document
Unlocking...