research and analytics

probability sampling

uses random selection to ensure members of a population have equal probabilities of being chosen

• considered more sound sampling methods compared to non-prob methods bc probability sampling results can be used to describe the target population as a whole

non-probability sampling

Techniques do not rely on random sampling of the target population but rather upon the researcher’s subjective judgement about how the sample should be structure, or upon access to the sample

• therefore the odds of any member of pop as a whole

confidence level

result of how confident researchers can be that if a study where replicated, the same results would be returned

margin of error

an indicator expressed as a percent range that estimates how much a statistic can fluctuate when using it to describe the true population. aka confidence interval or sample error

sample

a subset of a population, in market research, the population refers to the target of consumer/target market

• Researchers rely on a sample (subset of population)

• result from the sample is called a statistic

census

study that includes EVERY member of a population

• result is called a parameter

• IF a census collects data from every member of a population, it may be assumed that this is a superior technique and one researcher should strive to acheive, but often is not the case

simple random sampling

a completely random method of selecting subjects

• random number generator can be used to select those on the list for participation

stratified random sampling

sampling technique where subjects are split into mutually exclusive groups (called strata) before sampling is employed

• simple random sampling is then employed for those groups

quota sampling

involves gathering a sample in a proportional relation to predefined groups (cells) in the target population

• aims to ensure that there is enough data per demographic (age, gender, etc)

snowball sampling

research participants themselves suggest other members to participate in the study

• works well in populations that are difficult to find or identify

screening questions

used to qualify potential respondents to take part in a study

research panels

a collected group of potential respondents who opt in to participate in future research studies

• once opted in, panel members provide their contact info along with their demographic and psychographic info about themselves

• helps researchers target potential respondents who qualify for particular studies

• some companies maintain their own

• appeal is having access to a readily available sample od respondents from which to choose

• disadvantage: members are not exactly a random sample of the population

online/mobile survey distribution (survey distribution methods)

self administered surveys taken via the internet. disaggreated into 2 approaches

1. emailed or text invitations

   • advantages

     • respondents are affiliated with the topic in some way → helps response rate

     • surveys typically have to be brief, but since this group of people have a relationship with the survey topic, they can get away with longer surveys

     • ability to send targeted reminders

     • unique URL - company has respondents data like demographics

2. surveys accessed directly on websites

   • social media sites or elsewhere online when direct contact info is not given

surveys via mail - snail mail (survey distribution methods)

best for sensitive topics and when respondents need to gather info and contemplate their responses

useful when targeting a specific geographic location or when it is necessary to respondents in places with unreliable internet access

response rates low, costs are high

telephone interviews (survey distribution methods)

calling potential respondents and inviting them to partifcipate in an interviewer led survey

when contact is too general people do this

benefit is that the interviewer can clarify questions and probe

costs are higher, results lower

response rate

percentage of those who completed a survey out of the potential respondents on sample list

response bias

results from instances when a respondent answers a question in a way that misrepresents the truth

self-selection bias

occurs when respondents feel strongly about a survey topic and are, therefore, more likely to participate

non-response error bias

occurs when there is a significant difference between results of those who completed a survey and those who did not complete a survey

social desirability bias

occurs when respondents alter responses either to inflate their self-worth or provide responses that seem more socially acceptable than the truth

piping and screening

piping - a survey software system’s automatic insertion of text into a survey based on information from the sample list or based on responses to prior questions in the survey

screening - set of questions, typically delivered verbally or via written questionnaire, used to establish that a prospective participant is a good fit for a given research project

ethics in survey research

code of standard and ethics for market research and data analytics

• insight association is a membership organization representing companies, data analytics departments, and individuals working in the marketing, opinion, social research, and data analytics professions

• developed a code of standards and ethics

GDPR - general data protection regulation

• comprehensive data privacy regulation that was introduced by the EU to protect the personal data and privacy of EU citizens

• main goal is to give individuals greater control over their personal data and to simplify the regulatory environment for business operating within the EU

GDPR includes:

1. consent

2. data subject rights - GDPR grants individuals various rights over their personal data (erase, access, rectify, etc)

3. data portability - individuals have the right to receive their data in a commonly used and machine readable format and transmit data into another controller

4. data breach notification - orgs are required to report data breaches

5. privacy by design and default - orgs are encouraged to implement privacy measures from the beginning of the data processing lifecycle and to ensure that privacy is default setting

6. data protection officers - some organizations are required to appoint a data protection officer(DPO) to oversee GDPR compliance

7. cross-border data transfers - GDPR imposes restrictions on the transfer of personal data outside the EU to ensure that data is adequately protected

noncompliance to GDPR can resulting large fines

confidentiality

refers to not sharing individual responses with anyone outside of the research team, including study sponser

• researchers themselves have access to the individual data (contact info and responses) since raw data from surveys for instance is collected individually before being aggregated

• great care is taken to safeguard info so that it is only used by appropriate research staff

anonymity

refers to data in which even the researcher cannot identify respondents.

• Respondents are anonymous

• can be achieved by sending a generic link to the sample without identifying characteristics embedded

• creates some challenges: can’t send reminders because ALL respondents get the reminder even if they have already taken the survey

  • instead “blanket” reminders are sent → all potential respondents get a reminder

  • forgoes the ability to link responses to demographic information in the sample which could be used for ensuring a represnetice sample of the population for segmenting results

measure of central tendency

summary measure that helps describe a set of data in a single value

the value in a measure of central tendency represents the center of distribution

mean

sum of all the values divided by the number of responses

median

middle value in a distribution when the values are arranges in ascending or descending order

mode

most commonly occurring value in a distribution

range

difference between smallest value and largest value of set. provides an idea of how widely spread out the most extreme responses are

standard deviation

the standard deviation is a stat that measures the set of a dataset relative to its mean. the further the data points from the mean, the higher the standard deviation

• tells the researcher if the responses are concentrated around the mean or if they are scattered

• square root of variance

  • variance = subtracting value of mean from each data point then square those values, then add them up and divide them by N-1, calculate square root and thats the standard deviation

raw data

data that has not been aggregated or summarized in any way

data cleaning / clean data

refers to process of identifying and correcting errors, inconsistencies and inaccuracies in datasets

may involve

1. handling missing data

2. removing duplicates

3. standardizing formats

4. correcting inconsistencies

5. dealing with outliers

6. ensuring accuracy

7. coding categorical variables

8. transforming variables

9. excluding data from analysis

clean data = data that has no errors, is consistent, and accurate

top box scores

while mean scores are usually used when reporting results into interval questions (such as those with rating scales), percentages can be used as well to help summarize and compare the responses in a different way

• top box scores use the percentage of respondents who answered using the highest value in a rating scale

  • may be used instead of means to isolate a report the proportion of respondents who are very positive about a feature or a conceot

  • often used as a way to report the level of brand loyalty

• top 2-box scores combine percentages of those who answered the highest and second highst in a rating scale

word clouds

text-based visuals that give greater prominence to words that appear more frequently in a data set

• provides a visual summary of qualitative data, validates survey results, and identifies key themes in research findings

weighting data

refers to giving more or less power (or weight) to data in underrepresented or overpresented segments so that the data more closely matches the mix of the target population

• stat software can be used to weight certain cases in the sample to a desires level

coding open ended responses

coding open ended data = categorizing responses into themes. oftentimes there may be several different themes presented in one comment

• there are software programs that can help with the coding process but more often then not they have to be somewhat manual coding activities

involves developing “codes” to assign comments to themes

• to uncover the major themes, percentages are calculated for each code

• the denominator is typically the number of total comments gather for a particular question and the numerator is the raw number of comments that correspond to each respective code

• transforms qualitative data → quantitative