Coleta de dados

 9 erros comuns na coleta de dados de uma pesquisa científica

 

“Chamar o estatístico depois que o experimento foi feito é o mesmo que pedir a ele que faça um exame post-mortem – talvez ele consiga dizer do que o experimento morreu”.

Ronald Fisher, criador de técnicas estatísticas fundamentais como a ANOVA, sabia que muitos estudos morrem antes de nascer, na hora da coleta dos dados.

Tanto para experimentos de laboratório quanto pesquisas de campo, o planejamento e cuidado na obtenção dos dados é fundamental.

A maneira mais segura de evitar problemas pelo caminho é contar com o auxílio de um estatístico, mas isso não isenta o pesquisador de conhecer as boas práticas para coleta de dados, para evitar que termine com um material inutilizável para pesquisa.

Começando pelo básico, alguns erros comuns que provavelmente o orientador da pesquisa ajudará a evitar são os seguintes:

1. Não calcular o tamanho de amostra necessário

 

2. Não planejar a análise posterior a ser feita

 

3. Não ter objetivos de pesquisa claros em mente

Muitas vezes o pesquisador tem claro para si seus objetivos, mas não sabe como isso se traduzirá numa análise estatística. Por isso deve-se saber não só o que se quer investigar, mas como chegar ao resultado desejado, em termos estatísticos, sabendo quais testes usará.

Isso também envolve, com certeza, obter uma amostra que tenha poder de detecção satisfatório das medidas de interesse, e talvez realinhar seus objetivos se não puder obter o desejado.

 

4. Coletar amostra “por adesão”

Quem já participou de qualquer tipo de pesquisa é solidário à dificuldade de conseguir voluntários para responder questionários. Algumas pesquisas são ainda mais complicadas porque envolvem unidades amostrais difíceis de se obter.

Ainda assim, não se pode abandonar totalmente o critério de aleatoriedade e pesquisar apenas quem ou o que está mais disponível e próximo.

Tem se tornado corriqueira a pesquisa pela internet em formulários compartilhados por Facebook com os amigos e que com certeza tem um viés em seus resultados. Para quem preza a qualidade de seu trabalho, deve ter cuidado ao utilizar esse método.

 

5. Não treinar os coletadores de amostra

Assim como você não usaria uma máquina desregulada para medir as variáveis de interesse do estudo, não se pode colocar um pesquisador de campo aplicando questionários sem treiná-lo da forma adequada. Desde a abordagem do pesquisado até a ordem das perguntas, é necessário garantir um padrão. Caso esteja trabalhando sozinho em campo, policie-se para realizar as abordagens sempre da mesma forma.

 

6. Anotar os dados de forma imprecisa

Por descuido ou mau planejamento, pode-se acabar com informações menos precisas que o necessário. O exemplo comum é da pesquisa em que se pergunta a idade do pesquisado, e não sua data de nascimento. No primeiro caso, há maiores chances de respostas incorretas, por arredondamento, esquecimento ou mentira do pesquisado.

Queremos a informação o mais detalhada possível, pois sempre podemos arredondar ou agrupar os resultados depois. Também pode-se criar critérios que avisem de informações possivelmente erradas durante a coleta. Diversos outros exemplos podem se aplicar, o importante é antecipar-se  e estar atento para evitar incorreções.

 

7. Acreditar que quanto mais dados, melhor

Em termos gerais, mais informação não costuma atrapalhar o estudo, mas não se deve coletar todos os dados possíveis, deixando para depois a decisão do que fazer com eles (voltando ao item 2 dessa lista).

 

8. Depois da coleta pronta, querer que os resultados sejam avaliados em subgrupos muito pequenos da amostra

Algo que muitos pesquisadores iniciantes não tem ciência é de que nem sempre é possível avaliar a medida de interesse em qualquer subgrupo da amostra. Caso se tenha interesse em avaliar uma variável de interesse comparando homens e mulheres, por exemplo, é preciso garantir uma quantidade mínima de cada um desses subgrupos na amostra, o que não é possível de ser obtido pós-coleta.

 

9. Não cuidar dos dados originais

Seja em questionários de papel ou planilhas eletrônicas, os dados originais devem ser guardados até o final da pesquisa (e depois) pois não podem ser substituídos facilmente. Alguns pesquisadores também não sabem que informações resumidas não substituem os dados abertos, por unidade amostral, e a pesquisa ficará bastante limitada se perdê-los por algum motivo.

Se quiser, comente abaixo outros procedimentos básicos que podem ajudar seus colegas pesquisadores. Espero que o texto seja útil para os acadêmicos que estão começando a aprender a analisar seus dados e, na dúvida, contate um profissional para ajudá-lo.

Avaliação online

List of Online Survey Question Types

When it comes to optimizing your online survey, it’s important to consider the question formatting you’re going to use. While it might seem like a small detail, without using the correct format for your specific audience, you risk having your participants stop half way through, or abandon it half way through due to the complexity.

Multiple Choice (Select One)
The multiple choice selection is used for a one choice answer. This format is highly effective in terms of having your customer complete the survey, as it doesn’t require any form of additional answer. Radio buttons are the standard button used in multiple-choice questions. Your answer option orientation can be arranged vertically, or horizontally, depending on your preference.

Multiple Choice (Select Many)
The multiple choice can also be arranged in terms of a multiple answer question. Use the check boxes in the event your prospects might want to choose more than one answer. This works when you’re looking to get as many answers as applicable, and can be arranged vertically or horizontally.

Drop Down Menu
The drop down menu is generally used when there are extensive options available to your participants (for instance, country) and can be used when you want the subject to select only one option out of a massive group.

Open Ended Text Questions

Open-ended text questions allow your users the opportunity to write in their own text, depending on the question. This means your answers will become more complex, but subjectively could prove a higher ratio of insight.

Comment Box

The comment box is an open-ended text type question that allows the user to input his answer into the answer field. This would be the ideal place to put an “additional information” subheading, and give your users the ability to expand upon what might seem like a clear-cut answer.

Single Row Text

The single row text is ideal for short or one-line answers to a question. This would work well for online surveys needing a name, city or address.

Numeric (Freeform) Input
The numeric (free form) input allows a user to freely enter in any numeric (0-9) answer. This question is ideal for phone numbers, ages or date of birth.

Email Address
This question form is solely used to collect email addresses. Upon entering and submitting, only valid email addresses will be accepted.

Matrix Questions

A matrix question is a series of questions that pose the same answer scale according to a uniformed block. This can vary from Poor- Excellent, Disagree- Strongly Agree, or even Not Likely- Very Likely.

Multi-point Scales (Matrix Single Select)

This question is often used when testing a large group of participants, hoping to easily analyze and report data. It allows for consolidated reporting by providing a workable scale for easy interpretation. 

Check box / Multi-select (Matrix multi-Select)
Like the standard single select scale, the multi-select is used when testing a large group of participants. The multi-select matrix allows participants to choose any agreeable option with ease.

Spreadsheet
The spreadsheet matrix allows participants the opportunity to enter in their own (often numerical) data for specific questions. These questions work well when assessing specific financial amounts.

Rank Order
The rank order scaling questions allow specific order of importance to be established, based on the user’s input. These questions can be based on a specific attribute (like order of importance while eating in a restaurant) or basing the preference on an item as a whole in comparison with like objects. 

Constant Sum:
The constant sum questions help collect the ratio data in terms of online surveys. These surveys can permit expressive relative values or importance of options based on the option order. This can help in establishing overall importance factors, or deciding on whether or not an issue is worth addressing.

Contact Information

The contact information screen provides a collective set of data directly related to your customer. These contact forms will help maintain an organized compilation of all survey user data, and can be assessed based on criteria like last name, city or ZIP code. This can help you in terms of database and in geographical networking.

Avaliação

4 Measurement Scales Every Researcher Should Remember
 

One of the standard features offered by QuestionPro’s online survey software is a wide variety of scales that you can use to measure customer response.

At a first glance all the different scales that might seem similar and easily replaceable by each other. However, as you study them in depth, you realize the diversity of their natures and differences in their uses and their findings. There are over 20 different types of scales that are used by researchers in online surveys.  They can be categorized in two classes – comparative scales and non-comparative scales.

There are a number of factors you might consider when deciding on which scales to incorporate in a questionnaire and which ones to use while analyzing data. Some of the factors are:

• The type of data that is required from the respondent – ratio, interval, ordinal or nominal.
• How the information will be used once it is acquired.
• Number of divisions in the scale – odd or even.
• Types of statistical analysis methods to be used after data is acquired.
• The physical form of the scale – vertical, linear, horizontal, etc.
• Details to be provided in the scale as labels.
• Whether or not response to a question is mandatory.

Since non-comparative scaling techniques are easier and simpler to understand, we’ll introduce to you the most important four scales. You’ll be delighted to see how easy it is to understand and use them. Those who already know about it them are encouraged to comment on the post and let us know any tips that might further help our readers in using these scales.

1.     Graphic Rating Scale

A graphic rating scale, also known as a continuous rating scale usually looks like the figure drawn above. The ends of the continuum are sometimes labeled with opposite values. Respondents are required to make a mark at any point on the scale that they find appropriate. Sometimes, there are numbers along the markings of the line too. At other times, there are no markings at all on the line.

2.     Likert Scale

A Likert scale typically contains an odd number of options, usually 5 to 7. One end is labeled as the most positive end while the other one is labeled as the most positive one with the label of ‘neutral’ in the middle of the scale.

The phrases ‘purely negative’ and ‘mostly negative’ could also have been ‘extremely disagree’ and ‘slightly disagree’.

3.     Semantic Differential Scale (Max Diff)

A semantic scale is a combination of more than one continuum. It usually contains an odd number of radio buttons with labels at opposite ends.   Max Diff scales are often used in trade-off analysis such as conjoint.

MaxDiff analysis can be used in new product features research or or even market segmentation research to get accurate orderings of the most important product features. The SurveyAnalytics platform help’s you discriminate among feature strengths better than derived importance methodologies. Like other trade-off analyses, the analysis derives utilities for each of the most important product features which can be used to derive optimal products, using market segmentation to put respondents into groups with similar preference structures, or to prioritize strategic product goals.

You can have your respondents perform Forced-choice nature of the tasks, where in SurveyAnalytics MaxDiff can disentangle the relative feature importance in cases where average Likert-style ratings might all have very similar ratings.

4.     Side-by-Side Matrix

Another very commonly used scale in questionnaires is the side-by-side matrix.  A common and powerful application of the side-by-side matrix is the importance/satisfaction type of question.

First, ask the respondent how important an attribute is, then ask them how satisfied they are with your performance in this area.  QuestionPro’s logic and loop functions also allow you to run through this question multiple times with other alternatives that the respondent might consider.  This yields benchmark data that will allow you to compare your performance against other competing alternatives.

Here is an example of data from an importance/satisfaction question.  The importance rating is the line and the performance ratings are the bars.  With this type of data, you can actually see where your company needs to increase its efforts to more closely meet the needs of the customer.

While there are many online survey tools and online survey software to choose from, you’ll find that not all of them have these different types of scales available to them.

As you’re designing your survey, be sure to offer a variety of scales.  Using different scales in your survey will engage the respondent more fully and prevent them from clicking the highest, lowest or middle rating all the time.  Another benefit to using different kinds of scales in your survey is that each scale provides you a unique perspective on the data that you are analyzing.

Before designing your survey, review the different types of scales and question types inside of your online survey tool and be sure to pick the one that will best help you make your decision.