Sample Survey and Sampling Methods

Transcription

1 Sample Survey and Sampling Metods Course unit ECON MA/MSSc in Economics Dr. WM Semasinge Dept. of Economics

2 Tere are several important concepts relating to te sample survey and sampling metods. Population: 7/19/2017 Important Concepts of Sampling Population includes all members of a defined group tat we are studying or information collecting for data driven decisions. A population can be defined as te collection of all individuals or items wit te caracteristics one wises to understand. In statistics, population means te wole of te information wic comes under te purview of statistical investigation.

3 Population may be finite or infinite as te number of observations or individuals or items in it is finite or infinite. e. g. In a study on ouseolds average income in Sri Lanka, population includes all ouseolds in te country population is finite. In an investigation on te quality of a batc of a product of a certain firm, population includes all items produced by te firm in te given period of time population is finite. In te sampling process weter a population is finite or infinite depends on te ability to count all te individuals or objects separately practically.

4 In a study on te fises in te Indian Ocean, population includes all fises in te ocean - population is infinite. In a study on pressure at different points in te atmospere population is infinite Population may be ypotetical; eg. Outcomes of tossing a coin. A set of all possible or ypotetically possible outcomes of a statistical investigation. Te number of individuals/items of a population is denoted by N, and If te value of a certain caracteristic of i t individual/item is y i ten N = y 1, y 2,, y N

5 Sample 7/19/2017 A finite sub-set of statistical individuals in a population is called a sample. A sample is a scientifically drawn group tat actually possesses te same caracteristics of te population if it is drawn randomly. A sample drawn from a population provides valuable information about te parent population.

6 Sample size Number of individuals or items included in a finite sample is called size of te sample and denoted by n. If te value of a certain caracteristic of te i t individual/item of te sample is y i, ten, n = y 1, y 2,.,y n

7 Census vs. Sample survey 7/19/2017 Examination of eac and every individuals or items of a population is called Census or Complete enumeration. However, in many statistical investigations complete enumeration is impossible and impracticable. e. g. i. Estimating te average montly income of te people in India. (population is finite/100% inspection is impracticable) ii. Estimating te average eigt of a tree in te Sinaraja forest. (population is infinite/100% inspection is impossible)

8 iii. Estimating te average lengt of life of a electric bulb. (population is finite;100% inspection is impracticable) iv. Te cost of a Census in terms of time and money is very ig. Examination of a sample or part of a population to determine te population caracteristics is called sample survey.

9 Parameters and Statistics 7/19/2017 Various statistical measures suc as mean, median, mode, variation, standard deviation are estimated in statistical investigations. Wen tese measures defined for a population are called parameters and denoted usually by θ. Any statistical measure defined for a sample is called statistic and usually denoted by. Parameters are statistical measures based on population data wereas statistics are statistical measures based on sample data. ˆ

10 Statistical measure Population (θ) Sample (θˆ) Mean μ Variation σ 2 S 2 Standard deviation σ s Proportion P p Size N n x

11 Principle Advantages of Sample Survey 7/19/2017 Te principle advantages of sample survey as compared wit complete enumeration can be listed as follows: Greater speed Since only a part of te population as to be examined, data can be collected and analyzed witin less time period wit compared to te census. Terefore, result can be obtain more rapidly. Reduced cost Cost of te sample survey is less tan te complete enumeration. Altoug te cost per unit is generally greater of sample survey, total cost is expected to be muc smaller tan tat of complete enumeration.

12 Greater Accuracy 7/19/2017 Te results of a sample survey are muc more reliable tan tose obtain from a census because skilled and experienced personals, sopisticated equipment and tecniques etc. wic are important in a survey can be used sufficiently and careful supervision can be made in collecting and analyzing data.

13 Greater scope Sample survey as generally greater scope as compared wit census. In certain types of inquiry, igly trained personal and/or specialized equipment must be used to collect te data. In complete enumerations, it is difficult to afford suc personals/equipment sufficiently. Moreover, in sample survey, it is possible to ave a toroug and intensive enquiry because a more detailed information can be obtained from a small number of items/individuals.

14 In addition to tese advantages conducting sample survey is unavoidable in some instances suc as: i. Wen te population is too large or infinite. For example, trees in a jungle, fises in an Ocean, sand in a beac etc. ii. Wen testing is destructive. For example, testing te quality of milk, breaking strengt of loran bars, explosives, lengt of life of an electric bulb etc. iii. If te population is ypotetical. For example in coin tossing problem were te process may continue indefinitely, sampling is te only scientific metod of estimating te parameters of te population.

15 Limitations of Sampling 7/19/2017 Te advantages of sampling over complete enumeration mentioned above can be attained only if, i. Appropriate sampling tecniques is used, ii. Te sampling units are drawn in a scientific manner, and iii.te sample size is adequate. Sampling teory as its own limitations and problems wic may be briefly outlined as follows: 1. Proper care sould be given in te planning and execution of te sample survey, oterwise te result obtained migt be inaccurate and misleading.

16 2. Sampling teory requires te services of trained and qualified personnel and sopisticated equipment for its planning, execution and analysis. In te absence of tese, te results of te sample survey are not trustworty. 3. If te information is required about eac and every unit of te population, complete enumeration is compulsory. 4. If time and money are not important factors or if te population is not too large, a complete enumeration is better tan sample survey.

17 Te Principle Steps in a Sample Survey 7/19/2017 Te main steps involved in te planning and execution of a sample survey may be cited somewat arbitrary under te following eadings: i. Objectives of te survey and Formulating Hypotesis Te objectives of te survey sould be defined in clear and concrete terms. Tis is very important in planning te survey. Wat you want to learn? Wom you will survey and wat you will ask tem. Wat data sould be collected If te goals are unclear, te results will probably be unclear.

18 ii. Defining te population to be sampled 7/19/2017 Te population from wic sample is cosen sould be defined in clear and unambiguous terms. Defining of te population is not difficult wen sampling a batc of electric bulbs in order to estimate te average lengt of life of a bulb. Sampling from a eterogeneous population is not easy. Te rules must be set up clearly to define sampling units. Te enumerator must be able to decide in te field, witout esitation, weter or not to include a given unit in te population.

19 Te population to be sampled (sampled population) sould coincide wit te population about wic information is required (targeted population). Some times for reasons of practicability or convenience, sampled population is restricted tan te targeted population. Suc a situation, it sould remember tat conclusions drawn from te survey applicable only to te sampled population.

20 iii. Data to be Collected 7/19/2017 Te data to be collected depends on te objectives of te study. Wic data to be collected How to collect data Wo will collect te data Wen to collect te data It is important to verify tat all data relevant to te objectives are obtained and no essential data are omitted. All so, it sould ensure tat no irrelevant information are collected. Te practical metod to ensure tese is to produce an outline covering all information to accomplis te objectives.

21 Types of Data 7/19/2017 Primary data: Te data collected by researcer for is/er specific purpose and for is/er specific use. As tey are collected for te first time tey are original in caracter. Secondary data: Te data collected by oter researcers or institutions for teir purposes and available for te use of oters. Tey ave already been used and analyzed by someone else. Experimental data Observational data Cross section data Time series data

22 Metod of Collecting data (Primary) 7/19/2017 Tere are various measuring instruments and metods of measuring and collecting data from uman population. Survey Observation metod Questionnaire metod Interview metod Focus group discussions Participatory Rural Appraisal (PRA) Case studies Web survey

23 Te researcer sould cose te most appropriate metod among tese alternatives Accuracy aimed at and te cost involved are te important factors tat sould be taken into account. All tese metods ave certain advantages as well as disadvantages. Secondary Data Sources

24 v. Te Frame 7/19/2017 Before selecting te sample, te population must be divided into parts tat are called sampling units. Te list, map, cart or oter acceptable material wic construct including sampling units are called te frame. Te frame must, cover te wole of population not overlap, in te sense tat every element in te population belongs to one and only one unit. complete, in te sense tat it includes all units in te population, include te accurate information of te relevant units, up-to-date

25 vi. Selection of te Sample 7/19/2017 Tere are variety of plans by wic te sample may be selected. Sampling design is very important for te accuracy and reliability of te estimates. Selection of a proper sampling design is te responsibility of te researcer. In addition to te nature of te population, te relative cost and time involved sould also be considered before making a final selection of te sampling design. For eac sampling design, roug estimate of te sample size (n) can be obtain for te desired degree of precision.

26 vii. Te Pretest 7/19/2017 Before start te main field work, it is useful to try out te questionnaire and/or te field metods on a small scale. It will elp and allow, to improve te questionnaire or te data collecting metod, to plan te fieldwork, to identify te problems arising in various steps, to train te enumerates, to get an idea about te cost and time required for te main field work etc.

27 viii. Organization of Field Work 7/19/2017 Te success of a survey to a great extent depends upon te reliability of te field work. Tus, it is essential to trained te enumerators in locating te sample units, recording te measurements, te metods of collection of required data before starting te field work. It is very necessary to assign adequate supervisory staff for te inspection of field work. Plans must be made for andling non-response errors i.e. te failure of te enumerator to obtain information from certain sample units.

28 ix. Summery and Analysis of te Data Te analysis of data may be broadly considered under te few categories. 1. Scrutiny and editing of te data: Supervisory staff sould be carried out te quality cecking scrutiny in te ope of amending recording errors or at least of deleting data tat are obviously erroneous and inconsistent. 2. Tabulation of Data: Te data sould be tabulated employing most appropriate metod. However, before te tabulation it sould decide te procedure for tabulation of te data wic are incomplete due to non-response to te certain items in te questionnaire and were certain questions are deleted in te editing process.

29 3. Statistical Analysis: 7/19/2017 After te scrutinizing, editing and tabulating, te computations tat lead to te statistical estimates are performed. Different metods of estimation may be available for te same data. 4. Reporting and Conclusions: Finally, a report incorporating detailed statement of te different stages of te survey sould be prepared. In te presentation of te result, it is good practice to report te amount of error to be expected in te most important estimates.

30 x. Information Gained for Future Surveys More information tat we ave initially about a population, it will make easier to work out a sample tat will give accurate estimates. Tus, information gained from any completed sample in te form of data regarding te means, standard deviations etc. and te nature of variability of te population, wit te cost involved in obtaining te data serves as a potential guide for improve te future surveys.

31 Metods of Sampling 7/19/2017 A sample can be selected from a population in various ways. Different situation call for different metods of sampling. Tere are two metods of sampling: Sampling Metods Probability Sampling Non-Probability Sampling

32 Random or Probability Sampling 7/19/2017 Random or probability sampling is te scientific tecnique of drawing samples from te population according to some laws of cance in wic eac unit in te population as some definite pre-assigned probability of being selected in te sample. Te main probability sampling metods are: i. Simple Random Sampling ii. Stratified Random Sampling iii.cluster Sampling iv.systematic Sampling

33 Non-Random or Non-Probability Sampling 7/19/2017 Te metods tat sampling units being selected on te basis of personal judgment is called non-probability sampling. In tis metod, personal knowledge and opinion are used to identify te individuals/items from te population. It does not involve probability of selection. Te population may not be well defined. Tere are several non-probability sampling metods. Followings are te mostly used metods: Judgment Sampling Quota Sampling Convenience Sampling Snowball Sampling

34 Advantages of Probability Sampling It gives a representative sample even if te population is eterogeneous. Statistical measures (parameters) can be estimated and evaluated by sample statistic in terms of certain degree of precision. Since te estimates are unbiased, tey can be generalized to te population. It is used to draw statistical inferences. Matematical statistics and probability can be applied to analyze and interpret te data

35 Disadvantages of Probability Sampling Cost of sampling in terms of money and time is ig compared to te non-probability sampling. Non-response error is ig. Wen sampling frame is not sufficient, complete and up-to-date, te sample does not reveal te real situation.

36 Advantages of Non-probability Sampling Results can be taken witin less time period. Tus, tis tecnique is most popular in market researces. Non-response error does not arise. Cost is less tan to te probability sampling Scientific knowledge is not required.

37 Disadvantages of Non-probability Sampling Since te selection of sample units depends entirely on te discretion and judgment of te investigator, te sample may suffers from drawbacks of favoritism and nepotism depending upon te beliefs and prejudices As suc, te sample may not population. Statistical measures are not valid. Sampling teory cannot be used to test te statistical reliability of te estimates. Findings cannot be generalized be a representative for te

38 Sampling and Non-sampling Errors Tere can be discrepancies in te statistical measures of population, i.e. parameter and te statistical measures of sample drawn from te same population, i.e. statistic. Tese discrepancies are known as Errors in Sampling. Errors in Sampling are of two types: i Sampling errors ii. Non-sampling errors

39 Sampling Errors Sampling errors ave teir origin in sampling and arise due to te fact tat only a part of population, i.e. sample as been used to estimate population parameters and draw inferences about te population. Sources of sampling errors: 1. Faulty selection of te sample. 2. Substitution 3. Faulty demarcation of sampling units. Sampling errors is inversely proportional to te sample size. S.E S.Size

40 Non-sampling errors Non-sampling errors preliminarily arise at te stages of observation, collecting, processing, tabulating analyzing of data and publising te result. Tus, non-sampling errors may arise bot in sample surveys and complete enumeration. Sources of non-sampling errors 1. Faulty planning or definition 2. Response errors i. Response errors may be accidental. ii. Prestige bias iii. Self-interest iv. Bias due to interviewer v. Failure of respondent s memory

41 4. Non-response biases. 5. Compiling errors. 6. Publication errors. 7/19/ Due to te negligence and carelessness on te part of investigator. Non-sampling errors are likely to be more serious in a complete census as compared to a sample survey. Non-sampling error tends to increase as te sample size increases. Quite often, te non-sampling error in a complete census is grater tan bot te sampling and non-sampling errors taken togeter in a sample survey.

42 Simple Random Sampling (SRS) SRS is a tecnique of drawing a sample in suc a way tat eac unit of te population as an equal and independent cance of being included in te sample. Wen tere is N units of te population, eac unit as 1/N probability of being selected to te sample. Let us suppose tat a sample of size n is drawn from a population of size N. Tere are N C n possible samples. SRC is te tecnique of selecting te sample in suc a way tat eac of N C n samples as an equal cance or probability [p = 1/ N C n ] of being selected.

43 Selection of a Simple Random Sample Two metods of drawing a Simple Random Sample: i. Lottery system ii. iii. Using a random number table Computer based selection Available random number tables are: i. Trippet s Random Number Series, ii. Fiser s and Yale s Random Number Series, iii. Kendall and Badington Random Number Series, iv. Rand Corporation Random Number Series,

44 Caracteristics Definitions and Notation 7/19/2017 Te properties tat attempt to measure and record for every unit comes into te sample is called caracteristics. Population N Sample Total Y yi y1 y2... y N Yˆ y y1 y... i1 n i1 Yˆ Ny for Y i 2 y n is an unbiased estimater Mean N yi i y y N N Y 2 y n y n yi i y y n n y n

45 Sampling Proportions and Percentages Some times we wis to estimate te total number, te proportion or te percentage of units in te population tat posses some caracteristics or fall into some defined class. Number of unemployed person of a country Percentage of ouses tat does not ave electricity Proportion of people falls below te poverty line Every units in te population/sample falls into one of te two classes, say C or C Te classification can be done by including yes or no questions into te questionnaire.

46 Notation Number of units in C in 7/19/2017 Proportion of units in C in Population Sample Population Sample A a P = A/N P = a/n Unbiased estimate of A is Aˆ Np Variance of sample proportion (p) V ( p) PQ n N N n 1 Were Q = 1 - P Variance of Aˆ Np V ( Aˆ) N 2 PQ n N N n 1

47 Merits of SRS 7/19/2017 Merits of te SRS over te oter sampling metods, apart from te broad advantages of probability sampling metods are: 1. Since te sample units are selected at random giving eac unit an equal cance of being selected, personal bias is completely eliminated. Tus, SRS is more representative of te population as compared to te judgment or purposive sampling. 2. Sampling procedure is simple compared to te oter probability sampling metods. 3. Cost in terms of money and time is low compared to te oter probability sampling metods.

48 Drawbacks of SRS 1. Te selection of a simple random sample requires an upto-date frame. 2. Administrative inconvenience. 3. If te population is eterogeneous te sample may not be a representative sample. 4. For a given precession, simple random sampling usually requires large sample size as compared to stratified random sampling.

49 Stratified Random Sampling Population of N units is divided into subpopulations of N 1, N 2,.N L units considering a caracteristic of te population. Tese subpopulations are non-overlapping (mutually exclusive), and togeter tey comprise te wole of te population (exaustive). N N... N L Te subpopulations are called strata. 1 2 After determining te strata, a sample is drawn from eac strata independently. Te sample size drawn from te population is denoted by n, and te sample sizes witin te strata are denoted by n 1, n 2,, n L. Wen a simple random sample is taken in eac stratum, te wole metod is called stratified random sampling. N

50 In a study relevant to te students in a university stratification can be done based on te faculties, subjects, departments, etc. In a survey on te business firms, tey can separate into subpopulations based on te turnover, number of employees, etc as small, medium, and large. In a study on te ouseolds in a city, tey can be stratified taking into account te income levels suc as ig, middle and low income. In a study on agricultural farms, te factor for stratification may be te size of te farms.

51 Notation Te following symbols all refer to stratum N = total number of units n = number of units in sample y i = value obtained for te i t unit N W = stratum weigt N n f sampling fraction in te stratum N Y y N yi i 1 true mean N n yi i 1 sample mean n 7/19/2017

52 true variance 1 ) ( N i i N Y y S variance sample 1 ) ( n i i n Y y s L n S N st f n S W n N N y V ) (1 ) ( ) ( Variance of te estimated L L st y W N y N y 1 1 Mean of stratified sampling y st

53 Determination of sample size and allocation among strata In stratified sampling i. Decide te size of te sample ii. Planning to allocation among eac strata Two metods of allocation: i. Optimum allocation ii. Proportional allocation

54 Optimum Allocation Under te optimum allocation te determination of sample size and te allocation among te strata are guided by two principles. i.e. so as to: 1. Minimize te V y ) for a specified cost of taking te sample, ( st 2. Minimize te cost of taking te sample for a specified value of V y ) Te simplest cost function is of te form ( st C c 0 c n Were, c 0 = overead cost c = a unit cost of te stratum n = number of units taken from

55 In stratified sampling wit a linear cost function in te above form, te size of n for te minimization of is: ) ( st y V n c S N c S N n c S W c S W n ) / ( / ) / ( / Estimation of Size of n Estimation of size of n (1) Wen te cost of sampling is given, above equation (1) must be substituted into te cost function in order to compute te sample size (n): ) ( ) / ( ( 0 ) c S N c S N c C n

56 If V ( y st ) is given as V ( y st ) V sample size n is n ( W S 2 V (1/ N) WS c ) W S / c

57 Proportional Allocation 7/19/2017 Wen te number of sample units drawn from a stratum is proportional to te size of te stratum is called proportional allocation. Number of units selected from stratum is n Tus, in proportional allocation eac stratum is represented according to its size. eg. If N = 10,000, N = 300, and n = 1000, ten N N (n) n ,

58 Advantages of Stratifies Random Sampling 7/19/ More representative: In an un-stratified random sample some strata may be over-represented, oters may under represented wile some may be excluded altogeter. Stratified sampling ensures any desired representation in te sample of te various strata in te population. 2. Greater accuracy: Stratified sampling provides estimates wit increased precision. Moreover, it enables us to obtain te results of known precision for eac of te stratum. 3. Administrative convenience: as compared to te simple random sample, in te stratified random sampling, te time and money involved in collecting te data and interviewing te individual may be considerably reduce and te supervision of te field work could be greater ease and convenience.

59 4. Sometimes te sampling problems may differ markedly in different parts of te population, e.g. a population under study consisting of, (i) literates and illiterates, or (ii) people living in institution suc as otels, prisons, ospitals, refugee camps, and tose living in ordinary omes, or (iii) people living in ill areas and plain areas. In suc cases, we can take tis different parts as different strata.

60 Cluster Sampling A random sampling metod Cluster sampling metod can be used, 7/19/2017 Wen tere are no proper sampling frame oter random sampling metods cannot be applied. Some times forming a sampling frame is difficult and costly wen te sampling units are dispersed trougout te large geograpical area. If te entire population is unclear or unknown, sampling frame cannot be formed. If te sample clusters are geograpically convenient Te clusters are 'natural' in a population

61 In cluster sampling, 7/19/2017 Te population is divided into N groups, called clusters. Te researcer randomly selects n clusters to include in te sample. Te number of observations witin eac cluster M i is known, and M = M 1 + M 2 + M M N-1 + M N. Eac element of te population can be assigned to one, and only one, cluster.

62 One-stage sampling All of te elements witin selected clusters are included in te sample. Two-stage sampling A subset of elements witin selected clusters are randomly selected for inclusion in te sample. Multi-stage sampling Elements are drawn randomly in several steps from selected clusters for inclusion in te sample.

63 Convenience Sampling A nonprobability sampling metod. 7/19/2017 People are sampled simply because tey are convenient sources of data for researcers. Te researcer selects people based on is convenience. Primary selection criterion relates to te ease of obtaining te sample. Cost of locating elements of te population Cost of locating elements of te sample Geograpic distribution of te sample Obtaining te interview data from te selected elements.

64 Popular in opinion surveys. e.g. A survey on te service of a bank, 7/19/2017 Unsystematically recruit individuals to participate in te survey Visiting a sample of business establisments tat are close to te data collection organization. Seeking te participation of individuals visiting a web site to participate in a survey Perception over te service of a government institution Consumers view over te quality of a batc of product Opinion over an election result etc.

65 Purposive sampling 7/19/2017 A nonprobability sampling metod Sample units select purposely based on a logical manner. Objective of taking purposive sample is to produce a sample tat can be logically assumed to be representative of te population. Tis is acieved by applying expert knowledge over te population to select a representative sample in nonrandom manner.

66 - An example for purposive sampling, Selection of a sample of universities in Sri Lanka tat represent a cross-section of Sri Lanka Universities. - Tis required expert knowledge of te population. Te caracteristics wic are important to be represented te sample. e.g. large, medium and small universities, Public and private universities. Identification of sample units wic meet te various caracteristics tat are viewed as being most important.

67 Snowball Sampling 7/19/2017 A nonprobability sampling metod Tis metod is useful wen te population is idden or cannot trace clearly due to te various reasons including legal and etical matters. Te only way of finding members of some communities is by asking oter members. e.g. Heroin addicts. AIDS patients About an undisclosed group/organization

68 Te first step of tis metod is to find one or few members in te given population by any metod. Tis step is known as first round. Next, ask eac of te first round members about any oters. Tis list form te second round. Next, ask eac of te second round member about any oters. Tis process is repeated number of rounds and stop wen tey give te same names over and over again.

69 In eac round calculate te percentage of new names entering into te name list. Tis percentage is ig in te first rounds and ten will drop sarply. Wen te percentage of new entrants drops to around 10 percent, ten te process stops. Tis will be te sampling frame. From tis list a sample can be drawn randomly.

70 Snowball sampling works well wen, 7/19/2017 no population list is available; members of te population knows eac oter easiest way to produce a list close to te wole population Disadvantages Requires a lot of work wen te population is large; Isolated people will not be included in te study

71 Quota sampling 7/19/2017 Quota sampling is a type of non-probability sampling tecnique. First, te population is divided into strata or identify te different groups of te population. Mostly, Gender, profession, age, social condition etc. are taken as stratification factors. Second, calculating a quota for eac stratum: quota means te number of cases tat sould be included in eac stratum. It will vary depending on te make-up of eac stratum witin te population. e.g. male and female are or etc.

72 Tird, Continue to invite cases until te quota for eac stratum is met Once you ave decided te number of cases you need in eac stratum, you simply need to keep inviting participants to take part in your researc until eac of tese quotas are filled. For example, suppose tat you are interested in comparing te differences in career goals between male and female students of te university of Kelaniya. Population: number of students say, Stratification factor: Gender Proportional number of male and female students relative to te population: 2:3 Sample: 100 Sample sould include 40 male students and 60 female students.

73 Advantageous of Quota Sampling Quota sampling is particularly useful wen you are unable to obtain a probability sample, but you are still trying to create a sample tat is as representative as possible of te population being studied. In tis respect, it is te non-probability based equivalent of te stratified random sample. Quota sampling is muc quicker and easier to carry out because it does not require a sampling frame and te strict use of random sampling tecniques. Te quota sample improves te representation of particular strata (groups) witin te population, as well as ensuring tat tese strata are not over-represented. Te use of a quota sample, wic leads to te stratification of a sample (e.g., male and female students), allows us to more easily compare tese groups (strata).

74 Disadvantages of Quota Sampling In quota sampling, te sample as not been cosen using random selection, wic makes it impossible to determine te possible sampling error. It is not possible to make statistical inferences from te sample to te population. Tis can lead to problems of generalization.

75

76 Metod Purpose Advantages Callenges Questionnaires, surveys, Wen need to quickly and/or easily get a lot of information from people in a nontreatening way. Can complete anonymously Inexpensive to administer Easy to compare and to analyze Can administer to many people Can get lots of data Can be adapted into many forms (online, paper, verbal) Many sample questionnaires already exist Migt not get careful feedback Question wording can bias respondent s answers Impersonal Doesn t always get te full story Interviews Wen you want to fully understand someone s Get a full range and dept of information Develop relationsips wit stakeolders Can take a lot of time Can be ard to analyze or compare

77 Metod Purpose Advantages Callenges Observation Focus Groups To gater accurate information about ow a strategy actually operates, particularly about processes Explore a topic in dept troug group discussion, e.g. about reactions to an experience or suggestion, understanding common complaints, etc. View operations of a strategy as tey are actually occurring Can adapt to events as tey occurs (first and information) Quickly and reliably get common impressions Can be efficient way to get muc range and great dept of information in a sort time Can convey key information about strategy Can be difficult to interpret seen beaviors Can be complex to categorize observations Can influence beaviors of strategy participants Can be expensive Can be ard to analyze responses Need a good facilitator for safety and closure Difficult to scedule 6-8 people togeter

78 Metod Purpose Advantages Callenges Case studies To fully understand or depict stakeolder s experiences in strategy, and conduct compreensive examination troug cross comparison of cases (if cases are comparable) Fully depicts stakeolder s experience in strategy input, process and results Powerful means to portray strategy to outsiders Usually quite time consuming to collect, organize and describe Represents dept of information, rater tan breadt.