NaimExplain

If you can't or won't wait the entire lag between the youngest and oldest ages in which you're interested? Answer: You can run a staggered hybrid design, instead. This design has lots of smaller lags embedded inside it, but with many overlapping specific ages, so that you can conduct lots of double-hybrid-like tests. If all of these tests show no evidence of cohort or time-frame effects, then you can safely compare the "youngest" data to the "oldest" data to get the desired difference. In this way, you can compare, for example, 30 year-olds to 55 year-olds in a study that only takes 5 years to run and has "defenses" against the two main threats to aging research.

What are The standard solution to the threats posed by cohort and time-frame effects? Answer: Use a Hybrid design. The logic of this approach is that the odds of the two different threats producing the same difference in the data is vanishingly small, so if you find the same difference in both types of comparison, you can safely conclude that the difference was caused by the difference in age (and not one of the confounds). Note that there are three specific versions of the hybrid experiment, with fanciest, double hybrid, being worth the extra effort.

Under what conditions can you safely ignore these threats? Answer: When you are studying the effects of aging over very short time-lags (10 yrs), such as a few years or less, then you really don't have to worry about either cohort or time-frame effects.

What is the threats to the internal validity of longitudinal design? Answer: Conversely, only one group of subjects is used in a longitudinal study, but order cannot be counter- balanced, and the world "gets older" at the same time as the subjects, so any difference in behavior across ages might be due to the changes in the world, instead, which is known as a time-frame or zeitgeist effect.

What is the threats to the internal validity of Cross-sectional design? Answer: The groups of subjects in a cross-sectional study are currently different ages, so they must have been born at different times. This means that they are members of different cohorts and, so, maybe any difference in their behavior right now is due to a difference between cohorts, instead of a difference due to current age.

The trick used to deal with the confounds 1-the cohort effect and 2- the zeitgeist effect is. Answer: Employ both approaches at the same time. If you find the same results in both cross-sectional and longitudinal research, then the odds of those results both being caused by their own unique confound is very low. For example, if the chance that what is found in a cross-sectional study is really caused by a cohort effect, instead of aging, is about 10%, and the chance that what is found in a longitudinal study is really caused by a time-frame effect, instead of aging, is also about 10%, then the chance of getting the same set of results using both methods could be as low as 1%, which is far below the cut-off for chance (5%) that was allow in psychology. On the other hand, if you don't get the same results using both methods, then you must be very very cautious as to how you interpret the results. If nothing else, you'll need to figure out which - if either - of the patter...

In general, what's the first and most important thing to check before agreeing that a certain set of data should be treated as a quasi-experiment? Answer: The most important question is whether the subject variable is really more stable than the data variable. This must be true for the set of data to be treated as a quasi-experiment.

Why is it not OK to divide your subjects in term of good vs bad previous-night-of-sleep and look for a quasi-experimental effect on current mood? Answer: Even though the previous' night sleep occurred before the current mood, there are too many third variables that could cause both to make this a candidate for quasi-experimental analysis (examples: previous-day's-events, including previous-day's-meals and -exercise).

Why is it not OK to divide your subjects in term of high vs low depression and look for a quasi- experimental effect on anxiety? Answer: Depression is no more stable than anxiety. The reversed-causation explanation is as plausible as what you seem to be interested in.

Why, in general, are quasi-experiment not threatened (any more than "real" experiments) by the third- variable problem? Answer: Like "real" experiments, the interpretation (and internal validity) of all quasi-experiments can be threatened by confounds. On the surface, it might appear that quasi-experiments are in serious trouble because no attempt was made to create equivalent groups. But when you think about it in terms of causation, instead of just confounding, you often find that all of the third variables that could be the real cause of both the subject variable and the data variable are actually aspects of the subject variable, itself.

Why, in general, are quasi-experiments not threatened by the directionality problem? Answer: Like "real" experiments, most quasi-experiments are not open to reversed-causation explanations because (a) the subject variable was caused a long time ago and causation can't go backwards in time, (b) it's very hard for a variable to be more stable than its causes, and (c) most subject variables are random and permanent.

Which is preferred (under what conditions)? Answer: The former is preferred, in general, because it always has equal-sized groups, so it has the best statistics for a given total number of people. But it's only worth the extra effort when the population isn't close to evenly split between levels of the subject variable of interest. Examples: handedness, since left- handers are much less frequent than right-handers.

What are the two ways to run a quasi-experiment (in terms of sampling)? Answer: There are planned quasi-experiments, where you sample equal numbers of people within each level of the subject variable, and there are ex-post-facto quasi-experiments, where you just take one big sample and split the people into groups after-the-fact.

What do quasi-experiments have in common with "real" experiments and how do they differ? Answer: Quasi-experiments are like experiments in two ways: in both cases, you have a labile measure providing the data and, in both cases, you have another variable that you think of as a potential cause of the data variable. (Note: in most cases, the SV in a quasi-experiment is qualitative, just like the IV in most experiments is often qualitative, but this need not be true in all situations, so I wouldn't really include it here.) Quasi-experiments differ from "real" experiments in that the researcher doesn't have complete control over the potential-cause variable.

What is the definition of a quasi-experiment? Answer: A quasi-experiment is a correlational study with one variable being a very stable "subject" variable and the other variable being a labile "data" variable.

What are the difference between a Quasi-Experiment and "Plain" Correlational Study? Answer: A "plain" correlational study concerns two equally-labile variables, such as depression and anxiety, while a quasi-experiment concerns one very stable variable and one labile variable.

Why does the relative stability of the two variables matter? Answer: It matters because it makes certain causal explanations of the entire pattern of data much more or less plausible than others.

What three questions should you ask yourself when trying to decide which approach to take - surveys vs observation - for a given correlational project? How do the answers to these question "push" you towards surveys or observation? Answer: First ask yourself what you're trying to measure. Surveys are good for unobservable things, such as attitudes, while observation is good for observable behavior. Then ask yourself whether reactivity is a serious threat and whether realism is very important. A "Yes" to either question would push you towards observation. Finally, consider the amount of work that's involved. This might push you back towards surveys.

In terms of the four types of validity, what's the advantage of observational studies over most other methods? In terms of the threats to the various types of validity, same question. Answer: Observational work usually has more external validity (or less of a need for external validity) when compared with either experiments or surveys. The reason for this: observational research has very high realism without any reactivity (assuming that you don't get caught).

What is the general rule that limits when and where you can conduct observational studies without the consent of the subjects? What implications does this rule have for each of the two ways of conducting observational studies? Answer: The general rule on conducting observational studies without consent is that it may only occur when and where there is no expectation of privacy. Therefore, naturalistic observation can only be done in public places and participant observation can't be done at all (without prior consent).