Research Methods Unit 2 Chapters 4 5 and 6 Concept o A construct that we drive by mutual agreement from our mental images Example A chair We all see an image of a chair in or mind although we all imagine a different chair they generally have the same properties that make it a chair o Conceptions The characteristics that we attach to a construct Example The chair Conceptions would be being able to sit on it it has a o Conceptualization back ect Process of trying to identify what a concept is Dimensions Example crime has multiple dimensions Indicators o Property crime violent crime drug use ect Just specifying what dimension your talking about leads you in the right direction Something that tells you whether the concept if absent or present o If this is there then the concept is there o Think of it like a checklist the more indicators you have the more likely you are to have something People do not need to agree on indicators like they do o Operationalization with conceptions Process of developing an operation definition of the concept Defining the concept precisely enough that you can go out and collect data SPECIFIC Measurement o The assignment of numbers to objects or events according to rules o Levels of measurements Discrete or continuous Must be one or the other Discrete means it has a relatively fixed set of values and has distinct categories Must fall into a set category Example sex marital status race ethnicity religion Continuous variables have an infinite number values Can always be reduced Example age temperature outside number of crimes you ve committed o Level or precision starting with the lowest Nominal variable A simple classification no numerical order Means only that there is a difference between categories Categories need to be exhaustive and mutually exclusive o Exhaustive Every person included must fall into a category Often add a other category o Mutually exclusive Everybody needs to fit into only one of the categories Discrete Example gender political party married or single Still has all listed above exhaustive mutually exclusive ect But you can now rank the order o We don t know how much the ranks are different Depends but can be discrete or continuous Example security levels at prisons Ordinal variables o Low security o Medium security o Max security Interval Zero does not mean zero Equal intervals between ranks Example IQ test if you got a zero it doesn t mean you have no intelligence just that you have low intelligence Example Temperature in degrees F Continuous Ratio Example how many times you ve been arrested income Continuous Includes all from above Zero now equals zero Reliability o Requires that your measure gives the same result every time the same thing is being measured Does not require something to be correct just constant Example if the clock in your car is always 5 minutes early it is reliable o Types of reliability 1 Stability Reliability over time Does the measure provide the same results when applied at different points in time 2 ways to measure this o Test retest method o Parallel forms Administer a test and then do it again Same test each time Assuming there is no change in the thing being measure your score should be the same both times Problems People can remember what they said the first time Where you come up with 2 different forms for measuring the same thing then administer both How to do this At time one group A would take test A and group B would take test B during time two that would switch group A would take test B and group B would take test A 2 Representative Reliability across subgroups or subpopulations Example Off on the same amount of wrongness between males and females or between different races ect 3 Internal consistency Reliability with multiple indicators o Whether those indicators are reliable o Want those indicators to all point to the same direction trend at least NOT in opposite directions How to do this Split half method o If you measured a concept with 10 questions you would split them in half and different people would get a different set of five questions You want to see if the same results are reached regardless of what questions they received How to do this Item to total correlation o We have our 10 items I combine them all into a total index can do this in multiple ways then I see if the individual items are related to the total idea o We want a high correlation which means the first item is highly related to the total and so on If item 6 is weakly related to the total it may show something different How to do this Inter item correlation o Checking all the items in comparison to the other items not the total Item one should be related to item 2 3 4 5 6 7 8 9 and 10 and so on How to do this Cronbach s Alpha o Takes each of these inter item correlation and gets the average o Ranges from 0 to 1 o In general we like things above 0 8 4 Equivalence Inter rater reliability Assess the degree to which different observers give consistent estimates Example interactions between a child and the bobo dolls Different people were watching compare how many times each person saw the child hit the doll We want a high equivalence o How can we improve reliability Conceptualize clearly Not just crime but number of arrests for a violent The higher the level the more likely your measurement crime Increase level measurement is to be reliable Multiple indicators to have some good ones Pretests and pilot studies The more indicators you have the more likely you are Pretest example instead of asking 500 questions just ask 10 so you can see if they are reliable Pilot studies example Ask all 500 questions but to a much smaller group of people Use established measure Doesn t guarantee that its reliable Training Replication Train experimenters helpers equally Validity o The extent to which the measure actually reflects the real meaning of the concept it is supposed to measure o A lot more difficult to establish concretely o Classical test theory T X e True value observed value error We want e 0 o But it will never be 0 o So we want as close to 0 as possible o Because there is always some error our observed value is always off a little bit o Ways of evaluating how valid something is Face validity Low basic type of validity Whether the measure makes sense Relies on common sense Criterion related validity 2 types o Concurrent Comparing the results of your measure to some other trustworthy measure When your measure is associated with a pre
View Full Document
Unlocking...