Overview of survey methodology - Part 1 : Preliminary identification of the target. In this first article, we will address the issue related to the goal of a survey. In doing so, you make sure to have gathered all the preliminary information available. Then, you have an overview of the work to be carried out.

Overview of survey methodology - Part 2 : Assessing various characteristics of a survey environment and probable results. The second step that will be discussed here is to try to assess some characteristics of the survey. This will enable you to identify the structure of the environment of your survey.

Overview of survey methodology - Part 3 : The questionnaire. The questionnaire is often the most convenient tool. You must understand that the role of a questionnaire is to be a support to gather data. The work of preparation must be fully completed before going further.

Overview of survey methodology - Part 4 : The issues related to the distribution of the questionnaire. For the practical creation of your questionnaire you must address things other then the various theoretical issues. You need to ask yourself question about your interlocutor in order to avoid reliable results.

Overview of survey methodology - Part 5 : Taking decisions related to sampling. The procedure of sampling, is the identification of the people that you must ultimately reach with your survey. In order to acheive a proper sampling, your will have to look at various factors that are discussed here.

Overview of survey methodology - Part 6 : various sampling strategies. What makes the sample such a significant tool is that it is supposed to accurately represent the population. It is a small scale model which must be as perfect as possible.

Overview of survey methodology - Part 7 : analyzing the data prior to the survey report. To analyze the gathered data is more than simply count the various answers gathered. An efficient analysis must enable the reader to pinpoint various elements of comprehension.

Several visitors asked me for some articles related to survey methodology. So, here is the first part of a series of articles on this topic. This series will help those who need to do a survey to have an overall picture of the appropriate procedure. Then it will be easier to have an idea of the work to be done. Since some topics were already discussed in other articles, you find in the text a few links to a related articles.

Before going any further, some preliminary remarks must be made about this article and the others which will follow. The reader should keep in mind that the information therein can be applied generally. Because of that, it cannot be possible to take into account every single situation. Thus, the surveyor will have to make the appropriate adjustments that applies to its particular situation. Also, since the steps presented are linked together, the reader will notice, that often, a same priority can be discussed in more than one steps, but with the appropriate modification that applies.

If you have questions about this article or others to come, do not hesitate to contact to me ether by e-mail or by using the page of comments on this site.

The first step is to identify the goal of the survey. It is a preparatory step where it is important to gather all the preliminary information available. It is not the time to start studying the subject. It is an overview of the work to be carried out that is the issue here. Thus try to gather, inter alia, the following elements:

• The source of the survey: letters and correspondences, document of invitation to tender, resolution of board of directors,...
• The reason for the survey: history of the problem, events and anecdotes related the situation to be studied,...
• The organizational context: structure of the organization which desire the survey, who will be your interlocutor and its role regarding the survey...
• The available resources : various lists people to be contacted, procedure of confidentiality, budget allocated, material and other equipment usable...
• The awaited result:
  • is this step of the exploratory type?
  • does one wish to measure the size of a problem?
  • does one seek to check the effectiveness of a particular solution to a problem?
• The human context:
  • does the survey have psychological or moral implications ?
  • can it be possible that the results of the survey can jeopardize the objectivity of the researcher?
  • will the results influence the continuity of the organization?
  • are the results likely to be unpleasant?
  • which is the relation between the researcher and the organization?

Always take time to write down your ideas. That will save to you much time. In doing so, you have a formal identification of some characteristics of the context, all that will prove useful in later steps. Moreover, that will help you to avoid surprises by giving you a more accurate assessment of the work ahead.

At the end of this step, you should have an overall view of work to carry out. You are able then to determine if you must engage you in the procedure of the survey. If your answer is affirmative, it will be easier for you to specify the conditions you need to carry out the survey.

This preliminary steep might prove useful if you have to present a tender or an offer of service since it enables you to consider :

• The size of the implied costs.
• The kind of collaborations you can have within the organization.
• The realism of the objectives to be reached.
• The probability of having "unpleasant" results for the organization.
• The level of knowledge the organization have about what you have to investigate.

You should now be ready to go to the second step which will be explained in the next article. The second step will be related to what needs to be evaluated before going further on a survey procedure. Those evaluations should not only be about the tendencies to be expected from the results, but also to the structure of a particular survey procedure and the adequacy of the available resources.

Your webmaster: Frédéric D'Astous

In the preceding article, we dealt with the issue of precisely identifying what trigger the need for a survey. Now, you should have a more precise knowledge of what you have to do for your survey. This knowledge should includes clues about the following elements:

• why you need to do this particular survey
• what method can be the suitable to gather the data
• what kind of information can be gathered
• the organizational context of the survey
• the context of the environment of research
• the size and the costs of the steps to be realized

All the elements above can be called "the context of the survey". It is now time to look at the circumstances regarding the parameters of the research. It is the subject of this article.

The second step that will be discussed here is to try to assess some characteristics of the survey. At this stage you have to put all together the results of the first stage, which had gives you an overview of the work to be done. Thus, we now have to address the following issues :

• aspect of the results,
• evolution that can follow the results,
• adequacy of the resources,
• dangers to choose the theoretically ideal method.

In what is related to the aspect the results, you will have to put you attention on some characteristics related to the survey. So, inter alia, we can identify here the following :

The characteristics of the information you need :

• do you need information of a personal or private nature ?
• can you anticipate the kinds of answers or behaviors that might be observable ?
• it there a probability that the needed information seems trivial for the respondent ?

The characteristics of the environment of the survey :

• is the social environment opened or closed ?
• is there a conflict (or a sympathy) between the organization and those you have to reach ?
• are the individuals you need to reach identifiable with a list of some kind ?
• do the individuals tend to gather at specific places ?

In what is related to the trends that might be revealed by the results, you will need to assess to level of knowledge you have about the characteristics of the information you want to obtain with the survey. The more you know about those characteristics, the easier it will get to further build the questionnaire. Furthermore, this will help to pinpoint problems in the methodology of your survey .

It is also here that you have to decide whether or not, it is possible and necessary to identify hypothesis. Usually, in any survey, there must be one or more hypothesis to verify. But in practice, it happens sometimes that such identification is not possible, nor desirable.

Then, there is the issue of the adequacy of the resources. Often, people are planing a survey with the idea to make it ideal. Although such expectation inevitably respects the methodological characteristics of perfect surveys, the real life test is likely to be difficult. It might even be a trap ! Take the example of an organization that wishes to do a telephone survey in a given sector.

Since the telephone lists have sometimes various problems, this organization decides to proceed with the random creation of the phone numbers to be called. Such sampling should be flawless since each telephone number will be the fruit of the chance. At that point, everything is fine. But, it is the use of those numbers that can became quite tiresome.

If the survey is to be done in a zone with strong density of population, the various telephone zones are used with a stronger percentage than the zones of low density. If the organization is in rural area, and that it proceeds by creation of random of numbers, it will be confronted with a high level of nonexistent phone numbers.

An apparently ideal method can thus reveal several inconveniences in practice, if it is badly adapted to the characteristics of the environment of the survey. In our example, we will note:

• an increase in the quantity of phone numbers to check,
• an ineffective use of the data-processing resources,
• a reduction of the proportion of valid numbers compared to those which are not,
• problem of motivation of the personnel charged to make the calls too often unfruitful,
• an increase in the costs of data-processing and salaries of employees.

Here, you will notice the importance to consider the adequacy of the resources in relation with the characteristics of the environment of the survey. The paramount issue here is to decide, whether or not, the methods considered are efficient. Do not hesitate to perform a simulation if it is possible. Such procedure will indubitably help you to locate the stumbling blocks that wait for you.

If you make an overview of the information you have following step 1 and 2, you should be able to have a precise knowledge of the following subjects :

• organizational environment and its resources,
• environment of survey and people who composes it,
• suitable method to carry out your medium,
• possible trends in the results and their structure.

It is now time to work with the questionnaire. It is what we will do in our next article.

Your webmaster: Frédéric D'Astous

In the second article of this series, I explained why it is important to assess the evolution of various characteristics related to a survey. Among these characteristics, the following ones were discussed:

• Evolution of the results
• Form of the results
• Adequacy of the available resources

Now you should have a precise idea of all factors related to your study. You should also be able to identify what you seek. It is now, the time to build your tool. The questionnaire is often the most convenient one. It is why it is the subject of this article. But, if you decide that another tool is more suitable for your needs, the same procedure can apply. Only the shape of the tool will be different.

First, you must understand that the role of a questionnaire is to be a support to gather data. Nothing more! It is thus imperative that all the issues of the preceding articles were addressed and solved in a convenient way. This work of preparation must be fully completed before going further.

Too much often, some are using the questionnaire as a means to gather ideas. In doing so, they hope to find clues which will enable them to go further down road. Such a perspective is a total loss of time and energy. It will never be repeated enough, if you do not have a precise idea of what you have to measure, the construction of the tool is premature.

Fortunately, to help you, there is a way of ensuring you that you have all elements in hand. You must schematize your problem. This procedure was discussed, in the article entitled: How to link the goal of a survey with the questionnaire To link questions with the goal of a survey. This procedure is a hierarchisation of the elements pertaining to your survey. At least you can count 2 hierachical levels, the dimentions and the indicators.

Dimensions constitute, to some extent, the areas which should be examined to have a keen vision of what you want to study. Each area is a set of related elements. You will name each area by using the characteristic common to all contained elements. It should be noted that some sources use the term: sub-dimensions. They are subsets of elements within a dimension. Other sources will also gather dimensions within more global sets that will be named : concepts. The later are kinds of "super-dimensions" centered on the priorities common to each regroupment of dimensions. Starting at the most global level, it goes as follow :

• The concepts
• The dimensions (*)
• The sub-dimensions
• The indicators (*)

(*) Those elements are essential

The indicators are the elements of the end of the chain. The wording must be as short as possible. This will help you to detect any inconsistancy in your vision of each indicator. Then, each indicator constitutes the goal of a question.

From there, you have to write down your questions. For now, use question of the open type. You will have to adjust the type of question thereafter. In the wording of your questions, make sure that you are not rechaping the indicator. Sometimes the wording can inadvertantly transform the indicator. This is obviously to be avoided. If such a transformation occurs, that may means your indicator is not precise enough. You might even try to measure two indicators in a same question.

Once your questions are written, try to put them in the same order as in your diagram of dimensions and indicators. Read each one, then ask yourself if you can arrange them in such way to obtain a logical sequence, from the first to the last. To start from the diagram dimension-indicators is the simplest manner. This gives you a logical layout that you will just have to adjust on the basis of what you know of the point of view of the respondant. If the respondant is able to perceive an overall logic, your level of answer will be higher. You must pay much attention to the sequence of your questions.

In addition to the effort of creating an understandable sequence of questions for the respondant, you must take care of the following issues :

• Look at the resources available and make sure that you have all you need to measure accuratly what you wish to measure.
  • For exemple, a bulky questionnaire will imply significant expenses. Do you have the money ?
  • Depending on what you wish to measure, you can be forced to change your choice of tool. You may discover you need a grid of interview. Do you have the resources to carry out such interviews?
• Avoid any possibility for the sequence of questions to suggest the content of the answers. Sometimes, in the sequence fo your questions, you might force the reactions of the respondant. The overall logic must only be used to help the respondant.
• Identify the efficient types of questions (open, half-open and closed) you have to use. It is there that you must pay attention to the caracteristics of the environment of the survey.
  • The caracteristics of the invironment gives you clues about the kind of information you can obtain from the respondants, and then, the appropriate type of questions.
  • These environmental caracteristics also inform you about the kind of support you must provide to your respondants. The more important the necessary support is, the less a questionnaire of survey is suitable.

Once that made, you have in your hands the questionnaire. At the same time, you should have made sure that a quetionnaire for a suvey it is the most efficient means to gatter the data.

It thus remains to attack you with the problem of the distribution of your questionnaire. This will be the subject of the next article. We will also talk about elements related to the presentation of the questionnaire.

Your webmaster: Frédéric D'Astous

Following the last article, you should have done a questionnaire that follows the various methodological requirements of what you want to study. But, for the practical creation of your questionnaire you must address things other then the various theoretical issues. Thus, if you do not ask yourselves any question about your interlocutor and then use your questionnaire just as it is, you shall have a problem. Only those who have the same level of knowledge as you have, of your subject, might be able to answer your survey. Your questionnaire is likely "to be specialized too much "!

However, even if the respondents are seldom specialists (in the theoretical meaning of this word) although they have a precise knowledge of what they live. But this "knowledge" is not organized the same manner as the methodical logic which enabled you to build your questionnaire. The "knowledge" of the respondent is based on personal experience. It is an empirical knowledge. And since there is a variety of way to make a life, the experience which come out of it is also very variable. It is thus necessary that the researcher pays attention to that and adapts his tool. The topics that are discussed here, are related to this concern.

To be able to perform the distribution of your questionnaire, you must take care of the following elements :

• The method of distribution
• The formatting
• The pretest
• The procedure of return
• The sampling (this will the discussed in our next article)

The method of distribution of your questionnaire is first of all related to the constraints resulting from the resources available. Each technique of distribution implies costs. It is also important to think about the quantity of questionnaires to be used. By balancing these two factors, you should be able to you to carry out an efficient equilibrium between the two. However, you will have to consider 2 other factors which can influence this equilibrium : the level of answers and then precision of the results.

With relatively small populations, the precision of the results will roughly vary according to the proportion of sampled people compared with the whole population. This is not a linear relation. In fact, the bigger the population is, and smaller the relation become. However, if you have a small sample of people (lower than 30-35), you will not be able to carry out a survey which respect the statistical laws, your precision will not be calculable. Also, the precision is no more related to the ratio sample / population when you have a large population. Then, it is the size of the sample which is proiminent factor. This is why, when you are doing a survey in large group, this ratio is of no importance. For more details, you can consult a resource in statistics. Moreover, the next article will discussed the issues related to sampling.

The level of answers will vary according to the method of distribution. In a general way, more the mode of distribution is personalized, more the level of answer increases. Thus, if your questionnaire is distributed by mail, your rate of answers should be lower than if you use the telephone to join your respondants. The personalized aspect of your questionnaire leads us to consider its formatting.

The formatting is always very important. But first of all, you will have to determine to who this formating is targetted. Indeed, the criteria to be considered are not always the same ones.

If your questionnaire is distributed by mail, this formatting is torgetted to your respondant.It will have to be personalized in some way. You will have to write your instructions in order that they look simple and precise. If you have sections, you will have to identify them along with the instruction that apply to it. Your respondant should be able to locate what you await from him. He must know what you wish to have from him. The more the respondant knows what you seek, the more your level of answer might increases.

If your questionnaire is done by phone, you will have to think about the person who carries out the phone calls. So, in addition to the instructions for the respondant, you will have to include additional instructions for those who will administer the questionnaire. Take care to visually distinguish these two types of instructions. You will gain in effectiveness.

If your questionnaire is used as a basis for an interview, you will have to think of its flexibility. In an interview, it sometimes happens that the things do not occur as planned. Thus, you will have to think of the degree of freedom that you wish to leave to the person who carries out the interviews and to manage the formating of the pages from this constraint. A simple means, can be the use of various paper colors for the various sections. You do not want your interviewer to lose himself in your questionnaire...

Now, your questionnaire should presented itself in a practical form. If you are convinced of that, you must put it to the test by means of a pretest. This procedure enables you to check if your questionnaire does not hide some problems. Often, the problems are related to the instructions and the choices of answers. Look also at the wording of the questions. To be sure that your pretest is reliable, you will have to apply this one to each type of people that your survey has to reach. Once this work made, your questionnaire is ready.

You might also think of a return procedure. Whatever it is by telephone or by mail, you must have a procedure for receiving the information or remind the respondant to send it to you. If this procedure involves your respondants, it must be simple for them. For example, in the case of a postal questionnaire, the presence of a pre-paid return envelope will be useful. If it is by phone, you will have to imagine a procedure to know when you have to call again if the respondant is not availlable at the moment of your call.

It is now time to do the sampling. This step is the identification of those you will have to reach with your survey. This topic will be discussed in our next article.

Your webmaster: Frédéric D'Astous

• the kind of information which you need to obtain.
• the kind of exchanges which you must have from these people.

And then:

• the kind of people from whom you wish to request the information
• a precise idea on how and where to find these people.

The first 2 criteria will enable you to identify how you will have to interact with the people who will answer your survey. These criteria should not be new for you, since similar elements have been considered at the time of the creation of your questionnaire. To allow you to determine this more precisely, I suggest to examine these criteria as a continuum with the 2 following poles:

Sometimes, you will need to interact in a more personalized way with the respondents. For example, this is the cases when you wish to obtain information on how people overcame a crisis. As a matter of fact, it is possible that you already chose to carry out interviews to gather your data.

On the other hand, it may be that you need information related to general topics. In this case, your study probably will be more effective if you are able to reach the biggest quantity of people that you can. It is then the distribution of the respondents, more than the identification of people with "relevant living experience" which will be important.

The last 2 criteria will enable you to decide how you can proceed according to the actual conditions of the environment of your survey. It is like to compare the actual opportunities offered by your environment with what you ideally wish to carry out.

You must know precisely what are the categories of people you will target. This identification sould have already been made, since you carried out your pretest. These people constitute your target. Take care to examine the results of your pretest to be sure that there is no type of people that were ignored. It happens sometimes that the results of the pretest force us to question various details, sometimes that will mean to target people you had not thought of initially.

It is then the time to confirm how these people can be reachable. You should already know which tools you need to reach them. Then it is the time to know if these tools are availlable to you. Here some tracks to be considered:

• Is there a list of these people, or these groups of people?
• Do these people "hang out" in a specific place? At a precise moment?
• Do these people know each other?

If you have a list, then you can do nearly anything. It is the ideal situation because each person can be identified precisely. You are able to make a probabilistic study by doing a random sampling of your respondents. Whatever you make interviews on the telephone, a self-administered questionnaire, or anything else is nearly irrelevant. Since it is the ideal situation, all the options are open for you.

However, the real life is made in such way as the ideal situation is not happening very often.

It may be that this is not individuals who are identifiable, but groups. So, it gets a little more complicated. If you have a large quantity of groups, a random technique is possible. If you have a small number of groups, you are likely to have to decide the relevance to work with such or such group. The ground is then slippery because you must precisely know your needs to make the proper decision. Your study is no more probabilistic. If you try to circumvent the difficulty by making a random selection inside each group, and build your sample from that, then you might have a problem. The conditions which prevail in each group are different. Seemingly, you will be able to make your probability computations, but these will not take into account the distortions caused by the fact that each group generates its own characteristics. The conditions of data gathering are not constant throughout the sample. Moreover, it may be that a group, because of its personality amplifies some characteristics while others can become irrelevant. Perhaps you will need these characteristics to perform your analysis. Therefore you cannot think of all the members of your sample as a monolitic structure. Then you need to know the particularities belonging to each group.

If these people are "hanging out" at a particular place, it is this place which can be used as a basis for your sampling. The relations with the people you may reach is likely to be more personalized. You may have to forget the probabilistic study since it might not be possible. The identification of each respondent is then an individualized procedure. On the other hand, if you require data related to "sensible matters", you gain in precision because it is by this individualized relation that the gathered information will be more accurate. The respondent will thus be motivated to give to you what you need.

Finally, if these people know each other, you can gain access to a network. A network behaves like a group. It amplifies some characteristics to the detriment of some others. On the other hand, you gain some advantages, since you are able then to carry out a gathering of data that is exhaustive. Indeed, a network is usually able to make the identification of all its members. Just like a group. Your data gathering might not be probabilistic, but you will be able to have reliable results.

At the end, you must make a comparison with the conclusions provided by our first pair of criteria. The rule of comparison could then be formulated as follows: the more the conditions of your environment of survey cause you to use a more personalized approach, the more you must give to the relation with the respondent a personalized aspect. Why? Quite simply because if you need the collaboration of a person in particular, you must be able to prove to this person that what you need is really important. It should not looks trivial. The actual conditions of a survey can force you to adapt the goal of a survey. Otherwise, the individual will not feel respected and then, your study will be much more difficult to realize. This point was underlined at the beginning of our serie of articles.

Your webmaster: Frédéric D'Astous

There are two families of strategies. There are random strategies and those which are not. The advantage of the random strategies, it is that they you can perform calculations related to the margin of error by the means of various mathematical methods. When a strategy other than random is used, it will be necessary to compensate for any possible disadvantages of the selected technique regarding the survey environment.

Also, keep in mind that it is completely useless to make probability calculations on a survey carried out with a nonrandom sampling. These figures will mean absolutely nothing. The margin of error cannot be calculated when using these strategies.

What makes a strategy a random one? Obviously, it is the fact that the respondents are identified randomly. It is also that the probability for each respondent to be selected is the same one for all! That means, if you decide to oversample a subgroup, you will have a small methodological problem to manage.

On the other hand, you will note that there are various methodological schools as well as a variety of characteristics in each survey. It is up to the researcher to make the choices who correspond with its needs and characteristics of the survey environment.

Most of the time the following strategies are efficient and will lead to relevant results. But in some "borderline cases", only the researcher will be able to determine if the practical application of a strategy is truly random. Indeed, the strategies which follow can be adapted in various ways. Never forget that, sometimes, reality makes us make choices which might be methodologically "unusual".

The family of nonrandom strategies

Accidental sampling: The particularity of this strategy is that it looks like a fully random strategy. But it is not a probabilistic one ! This is a source of frequent confusion about this technique. A practical example of this method, it is when one carries out interviews on the street by selecting a person at each 5 minutes. In this case, one forgets that people who walk the street have characteristics related to the activities of the sectors of this street. It is thus not representative of the population in general which is devoted to several other activities... in other places! Moreover, the types of people who walks this street vary with the hours of the day and even with the days. On the other hand, if you use various : sectors, hours of the day, as well as days of the week, you should obtain interesting results to analyse.

Sampling by quotas: This technique is interesting insofar as it tries to regulate the disadvantages related to the accidental technique. By knowing that our population is composed of such and such type of people, you may try to meet people of each type and probing those in proportion of what they represent in the population. The way to meet these people being often accidental, the disadvantages you will have to deal with, are similar to those of the accidental method. If you decide to do your survey in various places to fill your " quotas ", you will obtain more reliable results. Obviously, those places should be carefully choosen.

The sampling based on the reputation: In each group, there are people having more significant statutes. By meeting these people you will have access to information and knowledge on the group that ordinary people do not have. But if you think that the opinions of those "leaders" are the same of the group, you will be confronted with a particular difficulty. There is always a difference between the people having some prestige in a group and the opinions encountered in this group. On the other hand, if you seek to know the general characteristics of a population, and if you are not concerned with proportions related to each sector of opinion, this method can be of some efficiency.

The family of the random stategies

The typical random sampling: It is the ideal situation. You are the happy owner of a list and you can choose randomly the people on this list. No big concern in this case everything is nearly fine. Nearly fine since an average lists is not perfect. The biggest problems about lists are their sources. Lists have a tendency to privilege some sectors of a group compared to others.

Systematic sampling: This technique is a variation of the preceding one. The difference lies simply in the fact that instead of selecting people randomly, you select people at a precise interval. The result is the same one, the respondants have all the same probability to be selected. You can consider this technique as an equivalent of the preceding one. Some even doesn't care do take the trouble to make a significant difference between the two. A small detail at which you will have to pay attention is to choose randomly the first to be selected. If not? The first respondant of each list will be always selected and your sampling will not be random any more!

Stratified sampling: This technique is useful when you need information on each sector of your population. You need a list and you need to know the proportions of each subgroup compared to the population. You are able then to select your respondants in a random way. If you decide to increase the proportion of some subgroups, you have the advantage of knowing the proportion which these people represent in the population. Thus, it is possible for you to rectify your figures when it is time to produce results applicable to all the population. All in all, your sample thus includes random characteristics in the sense that, in each subgroup, the choices are made in a random way. However, because of the lack of linearity of the sampling rate throughout the sample, some will consider that this technique is not completely random. A manner of circumventing this argument would be to carry out an oversampling of the sections which have a particular interest. You might want to differentiate the two series of sampling when it is time to produce the results for the whole population. However, one can consider that if a sub-group is sampled in a way more precise than the sample itself, one does not cause a major problem while carrying out the relevant adjustments of proportions. But it can be more tricky when it is time to manipulate the results.

Cluster sampling: This technique is useful when you have a large territory to cover. One can consider that it is an alternative to the typical random technique. It is a question here of cutting out our territory in relatively similar zones, and choosing them randomly. One will normally carry out a random sampling in each one of these zones. An alternative will be to probe all the people of each sector . It will be acceptable insofar as the groups are comparable in size. You can also use this technique when you do not have a list of individuals but rather of the lists of groups.

As it is noted, doing a sampling is often a question of compromise. The researcher always have to deal with the particularities of its survey environment and those related to the researcher himself and the tools availlable to him. Moreover, one will notice that the methodological preferences can also be take into account. Nothing is simple in the country of the social sciences!

The next article will deal with the analysis of your results.

Your webmaster: Frédéric D'Astous

To analyze the gathered data is more than simply count the various answers gathered. An efficient analysis must enable the reader to pinpoint various elements of comprehension. Those element should be obvious for those who will consult the rapport.

Before carrying out this compilation, you have to look back on how your survey was actually carried out. Even if the greatest care were taken in order to make an objective study, it inevitably happens that events or factors come to disturb the procedure. Each one of these factors can distort your results. It is thus essential to have an overview of all possibilities to be sure that nothing deplorable had occurred. Among various ones, I suggest you to consider :

• Did a significant event occurred before or during the survey ?
• Were the deadlines respected ?
• Did all the respondents were contacted in a comparable way ?
• Did some resistance were observed, from all or a portion of the respondents ?

If every factors is carefully examined and each possible impact evaluated, then, your results should be considered as reliable. It is necessary to look at the impact of every harmful factors that could have distorted on your study and then, to be cautious throughout your analysis. Since the reliability of some of your results might be jeopardized, you should be shure to look at everything. Be particularly attentive to locate effects of amplification or minimization.

To help you in the analysis, there are various softwares. It is not possible here to give you all availlable option. However, here are some interesting ways of lookong for clues :

• The distribution of your results
• The rating of a caracteristic
• The correlations
• The sorting of data
• The recodification

The distribution results

Looking at the distribution of your results is not as seemingly simple as it seems. It seems simple because the interpretation of each distribution may seems obviousness. But to draw workable interpretation from a graph requires some experience. Probability law can sometime cause effects which are only seemingly significant. Here are some examples of curves that gives you clues for a better understanding your results. It is you, who have to deceide in each case, when can reasonably apply some to your situation.

The typical case. It is a distribution similar to the normal curve. It represents an ideal distribution according to the laws of probability. The majority of the phenomena tends to behave this manner.
The presence of 2 subgroups of respondents. Each subgroup is represented by a peak. It is like having two normal curves in same the graph. It is then useful to proceed to the sorting of the results in such a way to identify the characteristics of each group.
The presence of a group of undecided. The undecided ones tend to regroup around a central tendency. Here, our group of undecided is revealed by a peak in the middle value of the scale. This distribution can make you conclude to the presence of two subgroups. It is where the knowledge of the researcher will be useful. The undecided ones can have precise characteristics. If so, they behave like any subgroup. Or they might come from various categories of respondents.

The rating of a characteristic

By combining some results your can build a rating of a particular characteristic. For example, a questionnaire related to the stress in a work environment could include some questions about the events lived by a respondent. By combining the results of these questions you are able to rate the level of stress resulting from the environment of the respondent. To gather results so as to create a rating, you must allot a weight to each result of each question. Then, you combine your weights so as to obtain the value of "stress level" for each respondent. In the case of the stress, you can give a negative values to the situations being able to have a protective role. Several software will allow you to carry out varied combinations allowing to account for the most complex situations.

The correlation

Another interesting procedure will be to compare the evolution of two results, one compared to another. It is what will enable you to establish correlation's between the behavior of various variables. When you identify a link between two results, it will be useful to examine the values of the various statistical tests which accompany your comparison. These statistical tests, created by your software of analysis, enable you to draw conclusion on the significancy of a specific relationship, its form and its level of strength. This procedure with already examined in another article entitled: Correlations and causalities

The sorting of data

It is sometimes useful to be able to isolate a group of respondents. It is the time to use the sorting function of your software. This procedure particularly useful when will be used in conjunction with the preceding procedures. A sorting can be carried out on the results themselves or on other results the researcher might have carried out. Thus, a sorting would help you to know which cretaria to use to the identify the various subgroups of your sample.

The recodification of data

This procedure is done by taking the various results of a question and rearrange those to create a new variable. This procedure is usually used to regroup answers that have some similar characteristics. For example, you can think of a scale of income from which your regroup the various levels in order that it corresponds to specific levels of living (low, average or high class). One can use this procedure in conjunction with other to explore various aspects of your data.

Then, the report itself

You are now ready for the drafting of your research report. You should have sufficient information to work on. There is however of various other procedures to help you analysis your data. They can sometimes be useful you. It is a good idea to consult the documentation of the software you are using to analyze your data.

The goal of this last chronicle was to make a review of what it is possible to do in order to identify the significant elements in your data. The golden rule is to seek to identify the characteristics of your population as is revealed by your survey. It is for this reason that it will be necessary to pay attention to the significancy of each result. Good survey.

Your webmaster : Frédéric D'Astous.