Download Understanding Populations and Samples: Sampling Techniques and Methods and more Study notes Psychology in PDF only on Docsity!
SAMPLING PROCEDURES
When a psychologist asks a subject to learn a list of words they do this with the intent of learning about memory processes in general (i.e., to go from the particular to the universal ). Three Necessary Conditions to go from particular to universal
1. Going from particular to universal must be logically correct. EG: To say some swans are purple after observing only white swans violates inductive reasoning. 2. Data collected must be reliable (i.e., if one group performed better than another, need to know if this will happen under the same experimental circumstances, or if it was an anomaly). EG: You performed very well on my exams compared to your fellow classmates 3. Subjects should be representative of the group to be used in a universal statement. EG: schizophrenics – ahh, but what kind – chronic, acute? To meet these conditions it is necessary to know something about populations and samples.
SAMPLES and POPULATIONS
- A population is a clearly defined group of individual, events, objects, etc. EG: all humans, all smokers, all hospitals, all oak trees, all morons (politicians), etc.
- The key word is all , that is, all members of a population, although different populations may differ considerably in size. Research is Conducted Within Populations: 1. When a population is small it may be possible to sample from all of its members EG: school football team, congregation, etc. 2. Populations are generally large and it is not possible to sample all of its members. EG: automobile trade union, scientific society, alcoholics, etc. 3. In case of large populations, only a subset or sample can be dealt with.
TYPES of SAMPLES
1. Samples are a representative portion of a population chosen by some clearly defined set of procedures and attributes. 2. Statements regarding the population can be made on the basis of characteristics of the sample. 3. Thus, to fulfill the purposes of inference , the sample must reflect ( be representative ) of the general population. Basis of all Sampling Procedures - Random and Independent: 1. Every individual in the population has an equal chance of being chosen ( at random ). 2. Selection of one individual in no way affects the selection of any other ( independent ).
Sampling Strategies
1. Uncontrolled Sample: researcher has no control in selecting this sample(s). This type of sample is biased in the direction of more vocal respondents ( issue of self-selection ). 2. Haphazard Sample: sample with poor selection measures. 3. Simple Random Sample: sample selected from an entire population where selection is unbiased among individuals. Equal chance of being selected or selecting one individual has no effect on the selection of others. 4. Stratified Random Sample: sample containing subgroups in a proportion that matches the original population. Sampling is always random. 5. Contrasting or Cluster Sample: group selected by using clusters or groupings from a larger population. Not purely a random sample. 6. Purposive Sample: a non-random sample that is selected on the basis of some characteristic that it possesses. 7. Convenience or Incidental Sample: a non-random sample that is chosen for a particular practical reason, usually availability.
8. Probability Sample: Application of statistics to determine that any given individual will appear in the sample at a pre-agreed probability. 9. Systematic Sample: A probability sample but not a random sample.
Haphazard Sample:
1. Sample with poor selection measures 2. Haphazard samples can be almost worthless EG: Surveyor has control over whom to sample but uses haphazard methods of obtaining people. EG: A television station sends a crew out to interview 10 people on the street with instructions to include 5 women, 2 blacks, 3 teenagers, and 1 adorable little girl. EG: Literary Digest obtained respondents from telephone books and automobile registration lists. Predicted that Landon would win the 1936 presidential election over Roosevelt by a landslide. Overlooked the fact that during the Depression people who could afford telephones and automobiles were more likely to vote Republican.
Simple Random Sample:
1. Sample selected from an entire population where selection is unbiased among individuals. 2. Equal chance of being selected or selecting one individual has no effect on the selection of others. 3. Used when the population is relatively homogeneous with respect to the questions of interest. EG: Selecting 20 students from a class of 80. Each student assigned a number and selected using a list of random numbers.
Stratified Sample:
1. Random sample containing subgroups in a proportion that matches the original population. 2. Entire population is classified according to some scheme. EG: socio-economic level, religion, geography, etc. 3. Stratified Sampling should be done on the basis of a potentially relevant factor for the particular problem. EG: blood type of an individual responding to a political poll would be an irrelevant factor. 4. The number of people sampled in each category should conform to the proportion of that group in the entire population. EG: 50% big city, 35% small town, 15% farm, then sample should have same proportions. 5. Treats population as 2 or more separate subpopulations and creates a separate random sample of each. EG: One sample from a female subpopulation and one from a male subpopulation of 80 students. You want the sample to contain one-fourth of the 80 students and have the same sex ratio as the population (i.e., one-fourth females and one-fourth males). Given 48 females and 32 males in class, will require 12 females and 8 males. Number females from 1 to 48 and males from 1 to 32 and select them using random-number table. 6. After categories and proportions decided, then the individuals chosen within these constraints should be randomly and independently sampled. EG: Banning fraternities on campus; 25 greeks – support; 25 non-greeks appose. Looks even, BUT whole school - 120 greeks and 450 non-greeks. A proportional division of the categories reveal an overwhelming opposition. 7. Can over-sample a subgroup of the population to purposely include some group at a greater frequency than is represented in the population. EG: Opinions of whites and Asians on some matter in Hong Kong. Include same number of Asians and whites in the survey to get a reliable estimate of the attitudes of Asians as whites, even though whites constitute only 10% of the population. You would stratify on race and include 50% Asians and 50% whites in your random sample.
8. Either select randomly from the clusters or use the entire cluster.
Purposive Samples:
1. Non-random sample selected for a particular reason on basis of some characteristic that it possesses; must be representative of entire population. EG: Town voting for president of the United States 2. Can almost be considered to constitute a population EG: A researcher may survey the opinions of the presidents of several leading colleges about desirable changes in the college curriculum. The opinions of these people may be more valuable than those that would be obtained in a random sample of all college presidents.
Main problem with purposive sampling:
1. An error in judgment by the researcher in selecting the sample may influence the results. EG: A list of leading colleges composed by a researcher is more likely to contain the researcher's own college than would a list compiled by someone at a different college. 2. Lack of information or knowledge from your subjects about the problem being investigated. EG: the presidents of leading colleges might not know what the most desirable curriculum would be for students at the colleges that constitute most of the population of colleges.
Convenient or Incidental Sample:
1. Random and independent sampling of a particular subgroup within the population because that subgroup is highly convenient and/or available. 2. Similar to purposive sample; selects a desirable group of people but differs in that there is usually no data to suggest that the subgroup is representative of the total population. 3. Differs from a cluster sample because the subgroup is not more relevant – it is simply available. EG: A researcher may want to study effects of integration on social development in school children. Many appropriate schools to choose among, but it is much more convenient to study one in the researcher's own city. Even though such selection is not random, one would usually be willing to generalize the results to other similar schools and similar children. EG: Students enrolled in Introductory Psychology courses. Assumptions with Convenient Sampling: 1. The sample is assumed to be representative of the total population. This assumption is too easily made. EG: Not truly representative of the general population, but commonly accepted as so…inbred rats (only a morphologically similar); and again, first year Psychology students participating in experiments for course credit
Probability Sample and Concept of Random Selection:
1. Application of statistics to determine that any given individual will appear in the sample at a certain probability. 2. Researcher knows the probability that any given individual will appear in the sample. 3. As a first approximation, selection is random when it is controlled by chance alone. EG: Selecting a state lottery number. Authorities want to be sure that no one will be able to predict number better than chance. 4. Equal-probability-of-selection : Selection process is random if every member of population has same probability of being selected, and selection of one individual is independent of selection of any other. EG: Problem of ordering individuals or nonindependence among members of the groups. Bob and Carol attend a party with 3 other couples at which there are to be two door prizes awarded. Each couple's names are placed on a single slip of paper and one paper pulled out of a hat. If Bob’s name is chosen, so is Carol’s. Thus, Bob and Carol each had a 1 in 3
chance of winning, but because only both or neither could win, their selection was not random. If the host put all the names on separate pieces of paper and pulled two names , however, then selection would be random. Both of them would have a 1 in 3 chance of being selected, as before , but Carol's chance of selection would not depend on Bob's.
The Sampling Frame:
1. To take a probability sample of a population, it is necessary to define the population. 2. Defining a population necessarily includes or excludes certain individuals or groups. 3. Each individual that falls within the sampling frame is called an element. EG: Suppose you want to take a survey of 10% of your research methods class. The population in this case is the class. The class, however, contains some individuals who have not yet officially registered for the course or who will drop before the end of the term. You must develop a definition of the population for the purposes of the survey , and this may be different from the actual population. For example, you may define the population for the purposes of the study as those whose names appear on the official class roster as of a certain date. Any who have not yet registered will not be considered, even if they are attending class.
Systematic Sample:
1. A probability sample but not a random sample. 2. Although systematic samples are not random, they may be perfectly good for practical reasons. EG: Sample of 20 students from a research methods class of 80 students. First step, obtain class roster, then identify each element. You could use students' names, but this would require some way of randomizing their names. More convenient to work with numbers, so you would identify each element (student) by a number 1 to 80. If you choose every fourth name from the class roster, you would have a probability sample, because 25% of the class would have been selected. The sample would not be random, because those whose names were in positions 1, 5, 9, and so on, had a 100% chance of being selected, and everyone else had a zero chance. However, if you randomly chose which of the first four positions you started counting from, this method would meet the equal-probability-of-selection criterion, but it would still be non-random because very first choice would determine selection of all subsequent elements. EG: If there is some structure to the list, as we have seen, the results will be nonrandom. For example, a list of couples might be structured such that the man's name always followed that of his partner. Choosing every 10th person on the list would result in the selection of all men or all women. However, if the list has no structure to it, results will be as good as random in practice. Sampling in a Nutshell (6 Points) 1. Surveys may use haphazard samples, purposive samples, convenience samples, or probability samples. 2. Random sampling has a particular scientific meaning and requires considerable care to perform. 3. The sampling frame is the population that is available and actually sampled. Each individual that falls into the sampling frame is called an element. 4. Probability samples include systematic samples, simple random samples, stratified random samples, and cluster samples. Systematic samples use a probability rule for sampling that is not random. Simple random samples are feasible only with relatively small populations. Stratified random samples permit the researcher to ensure that various segments of the population are represented proportionately in the sample. Cluster samples are commonly used with large surveys. 5. Interviewing is a complex skill that requires considerable training. 6. The randomized-response technique, in which it is impossible to identify whether an answer is true of a particular individual, can be used to reduce the social desirability bias.
To test hypothesis an experimenter used two rats, each of which had just given birth to eight pups. One rat and her litter were placed in a large cage with ample space and objects to explore.
The second rat's pups were separated from the mother, and each placed in a separate cage. These cages were small, and only objects they could see or hear were the 4 walls and food dispenser.
After 5 months, both groups were tested in a multiple-T maze using food as a reward. After 20 trials all non- deprived pups were running the maze without error, but deprived pups were still making several errors. Latter group frequently froze and had to be prodded to move.
Concluded that sensory deprivation inhibits intellectual development such that deprived rats do not have intellectual ability to learn a simple maze.
Answer 3. Any difference between the two treatments could be due to the fact that the rats in the two treatments had different mothers. Within each treatment the rats are genetically similar, but genetically different across treatments.
Researchers frequently use a split-litter technique , with half the subjects put in one condition and the other half in the second.
It is also not clear as to whether sensory deprivation inhibits intellectual development, or whether it leads to fear. Rats in sensory deprived condition exhibited high fear responses (freezing), and their poor performance may be due to this rather than any impairment of intellectual development.
Some seem not to find this fear response, and perhaps deprived group has more area to explore in the home cage. One could alter the type of deprivation. One could have both groups receive some period of familiarization with test apparatus or with a like apparatus, which may occur over several days. This familiarization period would insure that the apparatus is not a strange environment" and thus diminish a fear response.
Case 4. During World War II hypothesis tested that punishment is more effective than reward for training people.
Task: identification of enemy and friendly airplanes. Subjects sat in front of a radar screen as silhouettes of enemy and friendly airplanes flashed on the screen in very short exposures (1 sec). As each silhouette appeared, subject had to respond by pressing 1 of 2 buttons—1 button marked "enemy" and other marked "friendly."
Each subject participated for 2 hours on 5 successive days. On the 1st day, after each stimulus subjects were told if they had been right or wrong in their identification. Starting on the 2nd day, subjects were randomly assigned to one of two groups.
In Group A were given 10 dollars after every correct identification but were not punished for a wrong identification. In Group B received electric shock after every wrong identification but received nothing for a correct identification. Procedure continued for days 3 and 4. On 5th (test) day subjects followed same procedure except that no reward, punishment, or information given to the subject.
Number of correct identifications per 100 silhouettes was considered a test of the effectiveness of each training method. As expected, there was some loss of subjects over the 5-day period; about 5% of the Group A subjects and about 35% of the Group B subjects had dropped out by the 5th day.
Results indicated that on the 100 test trials given to each subject on the 5th day, the mean number of correct identifications for Group A was 80, and differed significantly from the mean for Group B (92).
Concluded that the hypothesis had been confirmed and suggested that all training programs be based on punishment.
Answer 4. Here is a problem of subject mortality. It seems probable that the 35% loss Group B subjects were those who were performing poorly and getting shocked. The data collected on the fifth day may be biased, because it compares the best of Group B versus both the best and worst of Group A.
One design change to check for this would be to record performance scores on everyone each day. In this way one would know the performance level of those who are dropping out in Group B (and perhaps drop out a matched set in Group A).
To prevent mortality one could have used soldiers on an army base as captured participants. Need to resolve whether or not it's ethical to force subjects to continue attending sessions in which they will be shocked.
Another solution would be to pay subjects in both treatments a lump sum upon completion of the 5 sessions to insure that they will attend. But by doing this, are you not also rewarding the punishment group, which may confound your treatments?
One last change could include addition of a treatment which involves neither reward nor punishment, and one that contains both.
Case 5. A YMCA official in a small town wanted some evidence to prove that his program was valuable in training future leaders.
He went back to the group's membership records and got the names of boys who were active members in his program 20 years earlier and also got the names of some boys who were not YMCA members.
He compared the two groups as to present occupations, salaries, and so on and found that the YMCA group was doing much better.
Concluded that this was due to the influence of his program.
Answer 5. The problem here is that one can't be sure that it was the YMCA experience that determined later success. Perhaps the YMCA group came from wealthier and better-educated families, which would itself explain later success.
It may not be possible to correct the design as it stands, since you would have to go back to old records, which most probably are incomplete for the type of knowledge that is needed.
A better design would be to randomly select two groups of boys—one group gets YMCA experience and one group does not—and follow their careers. This design would eliminate any self-selection factors , although this is not usually how organizations operate.
Another alternative would be in the realm of sociological designs in which one tries to partial out other determinants such as socio-economic status, etc.
Case 6. Development of test to predict success of prospective lawyers.
Random sample of lawyers selected from list in Who's Who under the assumption that they would be successful lawyers.
At a large mental hospital, 10 clinical psych students each interviewed six new patients in one group for 1- hours. In another group, 60 patients were given a battery of standardized psych tests, and the test results were interpreted by 3 clinical psychologists who had several years of experience in interpreting these tests.
Each psychologist interpreted test results for 20 patients. Both students and clinicians asked to list each patient's major problems and assign patient to diagnostic category (process schizophrenic, reactive schizophrenic). Also asked to predict how long patient would be in hospital before improving enough to be released.
Results indicated that interviews were 67% accurate in predicting diagnostic categories and 22% accurate in predicting length of stay. The tests group was 83% accurate in predicting diagnostic categories and 65% accurate in predicting length of stay.
Concluded that interviews are of questionable value in both diagnosis and prediction of outcome and should be discontinued.
Answer 8. One criticism is that it has compared the best of the tests with both the best and worst of the interviewers.
Also problem of comparing interviews of inexperienced psych students with test interpretations of experienced psychologists.
Better design would be to make sure "interviewers" and "interpreters" are of equal experience in diagnosis of categories and also with respect to "length of stay" norms of that hospital.
Case 9. Are males are more creative than females? If so, does male superiority in creativity increase under conditions involving ego?
Design was a 2 x 2 factorial design in which one variable was sex and the other variable was high and low ego involvement.
Ego involvement manipulated by telling half the subjects that the task was a measure of intelligence and that their scores would be posted on a bulletin board (high ego involvement).
Other half of subjects told that experimenter wanted to test reliability of a task and that they should not put their names on the answer sheets (low ego involvement).
Test of creativity was an "unusual uses" test in which a person is given the name of an object (e.g., hammer) and has to write as many different unusual uses for that object as they can in 5 min.
25 males and 25 females participated in each of the two ego-involvement conditions. Males were members of a senior ROTC class, and females came from sorority pledge classes.
Two objects were used: (1) an army compass and (2) a monkey wrench. Subjects were given 5 min for each object.
Results indicated that mean number of unusual uses for two objects for males was 4.1 under low ego involvement and 7.6 under high ego involvement. Means for females were 3.2 under low ego involvement and 2.4 under high ego involvement.
Concluded that hypotheses were supported.
Answer 9.
Senior ROTC students may be brighter than sorority pledge classes, who are mainly freshmen. At least it is a reasonable assumption that seniors who have survived three years of the college may be (but not necessarily) more creative than freshman females.
The task objects, i.e., monkey wrench and army compass, are biased in favor of a male sample, since both are objects used by males more than females.
Corrective design would equate both females and males on college experience and perhaps intellectual ability.
Task objects should be as equally familiar to males and females (perhaps "pencil," "shoe," and "suitcase"). This, of course, assumes that the test is valid (which may be questionable).
Case 10. Researcher asked to conduct a quick survey in 3 large cities to find out what political issues or problems were important to voters.
Survey results were to be used by a political candidate in developing his campaign.
Researcher randomly selected names from phone books and called as many as he could reach between 9 A.M. and 5 P.M. on Monday and Tuesday.
Results were collated and presented to candidate with statement that they were a valid representation of atti- tudes of voters in 3 cities.
Answer 10. Definite bias in survey procedure in that the researcher only called between the hours of 9 A.M. and 5 P.M. One could assume that most respondents were housewives, househusbands, retired persons, or University students who would be home or either asleep at that time.
Issues of primary importance to them may not be representative of the total voting population.
One strategy frequently used is to randomly select phone numbers and call the chosen numbers until some response is obtained. Frequently the survey group will call only in the evenings and weekends and alternate asking for the man or woman of the house to balance out the sex distribution. Although these procedures may have some flaws (certainly you will miss people who don't have phones), they probably will give a fairly accurate picture in a short period of time.
