Total = 37 slides1Mid-Term Exams• Understanding your grade– The number printed on your answer sheet is the number correct (out of 27).– Quadruple the number correct to learn your grade (out of 100 points)– The mean score was 75.• 8 people got 100 (I rounded down to 100 for those who went over)• Two had perfect exams•How will your life turn out?– What’s possible: 300 + mid-term grade– What’s probable: (2 x mt) + ((2 x wkbk) [max = 200])•Total = 37 slides2Sociology 201:Social Research Design16. Experiments ITotal = 37 slides3Preview•Workbook assignments due: 8.2(40 pts)•Experiments– Review deductive model– Classical experiment–Internal invalidityTotal = 37 slides4Review deductive model--"traditional" image of science•Theory•Hypothesis•Operationalization•Observation: hypothesis-testingTotal = 37 slides5The logic of the experimental model• aims to avoid a number of weaknesses in day-to-day explanations•We’ll begin with a simple example to demonstrate the need for such logic and the procedures that go with itTotal = 37 slides6Why don't these statements prove that Vitamin C prevents colds?Vitamin C No cold?Total = 37 slides7“I took Vitamin C and didn't catch a cold.”Vitamin C No cold?Total = 37 slides8“My friend also took Vitamin C and didn't catch a cold.”Vitamin C No cold?Total = 37 slides9“My other friend didn't take Vitamin C and he caught a cold.”Vitamin C No cold?Total = 37 slides10“Seventy percent of those who took Vitamin C didn't catch colds.”Vitamin C No cold?Total = 37 slides11“Last year all of our family caught colds; this year we took Vitamin C and nobody did.”Vitamin C No cold?Total = 37 slides12What would you need for proof?Vitamin C No cold?Total = 37 slides13Anticipating experiments•Several of the comments so far point to the need for before and after measurements (pre- and post-)•Some comments reflect on the need for what we call control groups.Total = 37 slides14The Classical ExperimentPre-test Post-testStimulus(DV) (DV)(IV)ExperimentalControlGroupGroupTotal = 37 slides15ComparisonsPre-test Post-testStimulus(DV) (DV)(IV)ExperimentalControlGroupGroupABCDTotal = 37 slides16Problems of Internal Invalidity•Let’s suppose we want to improve prison morale•We decide we could do it with conjugal visitsTotal = 37 slides17Let’s suppose:•We measure morale among prisoners: average = 1/10•We run the program of conjugal visits•We measure morale again: average = 3/10•Can we conclude the program improved morale?Total = 37 slides18History•Something may happen in the outside world•Maybe the prison kitchen gets a new chefTotal = 37 slides19Maturation•May get tired, bored, wiser, older, etc.•Prisoners may get reconciled to being in prison•They may learn how to play the game betterTotal = 37 slides20Testing•Hawthorne Effect•Paying attention to prisoners’ morale may improve itTotal = 37 slides21Instrumentation•Pre- and post- measurements may not be comparable•May use two different questionnaires to measure moraleTotal = 37 slides22Statistical regression•Naturally drift away from extremes•Started with 1/10; couldn’t get much worse?Total = 37 slides23Selection biases•Control not comparable•Maybe asked for volunteers for the program•The others might not have wanted their spouses around•Might have offered program as reward to good prisoners•Or maybe use another (worse) prison as the controlTotal = 37 slides24Experimental mortality•People drop out•e.g., The sexually frustrated prisoners go over the wallTotal = 37 slides25Causal time-order•Improving morale might lead to a particular group being chosen for the experimentTotal = 37 slides26Diffusion or imitation of treatments•"Contamination”•“Control” warden might decide it was a good idea and start doing it tooTotal = 37 slides27Compensation•Feel sorry for controls•“Control” warden might start having “beer and pizza night”Total = 37 slides28Compensatory rivalry•Controls may work harder•“Controls” might organize to protest being left out and gain morale through their working togetherTotal = 37 slides29Demoralization• Controls may give upTotal = 37 slides30Discussion Example: • Students have trouble with statistics• Let’s say we think that giving them personal computers will improve their ability to learn statistics• Suppose we give computers to a class to use• At the end of the semester we give a test and the students average 90% on it.•Is this sufficient evidence? Why not?Total = 37 slides31Experimental ResultsPre-test Post-testStimulus20%90%ExperimentalGroupControlGroup20%70%Are you satisfied the computers made a difference?Total = 37 slides32Experimental Design•Dependent variable [Ability at statistics]–How would we measure it?Total = 37 slides33Experimental Design•Stimulus or Independent Variable [Personal computers]–Each student given a personal computerTotal = 37 slides34Experimental Design•Experimental and control groups–Pick a random sample of statistics classes–Randomly assign to two groups–Experimental classes will get computers–Control classes won’tTotal = 37 slides35Experimental Design•Pre-testing and post-testing–Use the standardized examination before and afterTotal = 37 slides36Discuss Internal Invalidity•Diffusion or imitation •Compensation•Compensatory rivalry•Demoralization•Statistical regression•Causal time-order••• History• Maturation• Testing• Instrumentation• Experimental mortality• Selection biases•Total = 37 slides37Next Time•Experiments II•Review Chapter