Chapter 10. Respondents and Sampling

1 AN EVALUATION GUIDE BOOK PAGE 81 Chapter and SamplingIn Chapter 9, we described several tools forcollecting data. Many of these tools, includingwritten surveys, interviews, focus groups, tests ofknowledge, and published instruments orassessments, involve asking questions ofrespondents. This Chapter will help you identifywho your Respondents should be and how manyyou are Respondents ? Respondents are the group of people fromwhom you will collect the information you need tohelp you answer your evaluation questions. Theyare the people you will interview or survey or whowill participate in your focus do Respondents differ from our targetpopulation?As described in Chapter 6, your targetpopulation is the group of people you are trying toreach with a particular strategy or activity.

2 Forexample, the target population for a community-wide information campaign would be the entirecommunity. The target population for a parents hot line would be parents. You will be collectingyour information about your target respondent is the person who will provideyou with information about the target , a respondent is a member of your targetpopulation. For example, if your target populationis the parents who attended a parenting class, thenyou will want to collect information from , the respondent will not be part ofyour target population. For example, you mightwant to learn about whether parents arevolunteering in their child s classroom. The groupyou are targeting is parents. It is possible to askparents about their volunteering directly, but youmight choose instead to ask teachers aboutwhether more parents have been volunteering ,since teachers may be more objective.

3 If you askteachers for information about the parents, thenthe teachers are the Respondents and the parentsare your target do we identify who our Respondents shouldbe?You want to obtain your information from themost direct and reliable sources whether individuals from the targetpopulation are the best ones to answerinformation about themselves, or whethersomeone outside of the target population wouldbe a better respondent. In the example just given,you might recognize that parents are the mostdirect source of information but decide thatteachers may be more objective in theirassessments of parent participation and easier foryour collaborative to contact. For many strategies,children are the target population. For some typesof information, the child might be the bestrespondent.

4 For example, children are in the bestposition to answer questions about feelings orconcrete experiences, such as what they had forbreakfast. For other types of information, thechild s teacher or parent might be the betterrespondent. For example, parents may be the bestsource of information about behavior at home orabout the child s past should we decide on our Respondents ?Decide on the Respondents when deciding onan evaluation method. For example, if yourevaluation question is whether children are awareof the importance of brushing their teeth, thepeople who can best answer this question are thechildren themselves, so you would want to be surethat the method you choose is an appropriate toolfor collecting information from children. You willneed to use words that children will work better than questionnaireswith young children.

5 If questionnaires will be usedwith older children, they need to be written at anappropriate grade level. Direct observation isanother method to consider when you need tocollect data on children. Remember that whencollecting data from children, it is almost alwaysnecessary to obtain parental many Respondents do we need?The more Respondents who give youinformation, the better. Of course, you have veryreal constraints in terms of how much time andmoney you have to collect information. Oneimportant issue you need to decide is whether youcan collect information about each member ofyour target population or whether you need toobtain data from a smaller group. If your targetpopulation is small, it is often possible to collectdata on all or almost all of the people in yourtarget population; then all of your targetpopulation are Respondents .

6 If the targetpopulation is large, it is usually necessary to collectdata from only some of them. When only some ofthe people in your target population are yourrespondents, this subgroup is called a A G E 82 AN EVALUATION GUIDE BOOKWhat is a sample?A sample is the group of people from yourtarget population that you will be collectinginformation from if it is not possible to collectinformation from all of good sample is representative of thepopulation you are studying, , thecharacteristics of your sample are the same asthose of the population from which it is other words, it is made up of the sameproportion of men and women , young and old,wealthy and impoverished, Democrat andRepublican, etc. Your sample will have a goodchance of being representative if every member ofyour target population has an equal chance ofbeing selected as part of the sample and youchoose the people in your sample randomly.

7 Thisis called a random do we select a random sample?You can select a random, representative sampleby following these steps:1. First, make or obtain a numbered list ofyour target Decide on the size of your sample (seediscussion on sample size below). Keep inmind that you will probably not obtainresponses from everyone in your Divide the size of your target population bythe number of people you need in yoursample. Round to the nearest Using the number you arrived at in Step 3,count down your list, circling selectedpeople. If your number is 4, for example,select every fourth Make a list of people chosen. This list isyour you cannot obtain a list of your entire targetpopulation, there are other Sampling techniquesthat you can use. You may wish to consult theresources listed in Appendix A or contact a localcollege, university, or research group large should our sample be?

8 If you have the resources, we recommend thatyour sample be 10% to 25% of your targetpopulation, but no fewer than 30, the smallestnumber generally considered adequate to , the larger your target population, thesmaller a percentage of that population yoursample needs to you are interested in comparing anysubgroups within your sample, you will need atleast 30 of each group. For example, if you areinterested in comparing teen parents with olderparents, you will need at least 30 teens and 30older parents. Subgroup analyses make yoursample size needs grow considerably. Continuingthe example, if you wanted to compare thefindings for Latina teen mothers and Anglo teenmothers, you would need at least 30 of each teengroup, for a total of 60 teen , the general rule is to make the sampleas large as you can afford in terms of time andmoney.

9 A large sample will best represent thewhole group. The smaller a sample becomes, theless your results will reflect accurately those youwould obtain if you had been able to studyeveryone. However, a small, truly random samplewill give you more accurate information than alarge sample that is nonrandomly chosen, becauseof the possibility of Sampling is Sampling bias?Bias is a systematic error that can prejudiceyour results in some way. When a sample is biased,your Respondents will not provide you with a truepicture of what is happening with your targetpopulation as a whole. For example, if, instead ofchoosing a random sample from a list of yourentire target population, you decided to pass outyour surveys on a street corner one Sundaymorning and the corner you were standing on wasclose to a neighborhood church, it is likely thatyou would end up with a sample that included ahigher percentage of people who attend churchregularly than is true of the neighborhood as awhole.

10 This group is likely to have views that differfrom those of people who do not attend views would be overrepresented inyour sample and bias your bias is a particular concern when yourtarget population includes people with lowincomes. If you use the telephone book to chooseyour sample, for example, it is possible that yoursample will underrepresent people with very lowincomes because they cannot afford a Respondents are often more difficultto reach than more affluent ones because theiraddresses and telephone numbers change morefrequently. It is important to keep this challengein mind when designing your evaluation will we know whether our sample is biased?Before you collect information fromrespondents, it is important to think about whichdemographic characteristics of your respondentsmight have an important relationship to their AN EVALUATION GUIDE BOOK PAGE 83answers.