Help
Options
Didn't know it?
click below
click below
Knew it?
click below
click below
Don't Know
Remaining cards (0)
Know
retry
shuffle
restart
0:00
Embed Code - If you would like this activity on your web page, copy the script below and paste it into your web page.
Normal Size Small Size show me how
Normal Size Small Size show me how
1d UNPRINTED
Statistics
| Question | Answer |
|---|---|
| Variable | Anything that can be measured that may change |
| Data can be collected from different sources | Experimental (laboratory, field, natural) Simulation Questionnaires observation reference census population sampling |
| Experimental | Experiments measure how one variable changes when you change another variable. The explanatory and response variables should be identified before the experiment. |
| Laboratory Experiment | Explanatory and extraneous variables can be strictly controlled. Very reliable and valid. However, people may not behave naturally. |
| Field Experiment | Done in real-life environment. Useful to know how something behaves in real world, or when using a lab is impractical. The explanatory variable is controlled, but there is less control over extraneous variables so there is lower validity and reliability. |
| Natural Experiment | If explanatory variable is too hard to control in field experiment, values of explanatory variables aren’t changed and the existing values’s effects on the response variable are seen instead. Reliability and validity are low as there is minimal control. |
| Simulation | Prediction of what might happen, based on previous data or known probabilities. Useful when collecting real data may be time-consuming, expensive or impractical. To simulate data, use random numbers. More times repeated, the more reliable the estimation |
| Methods of getting random numbers for simulation | Calculators: Ran# or RanInt function Computers Dice or cards |
| Questionnaires | Written list of questions |
| Questionnaires pros | Quick and cheap Well written shouldn’t be biased No pressure, so likely more truthful Lots of ways to distribute them |
| Questionnaires cons | Distribution can lead to bias Ignorable, so many non-responses Misunderstood questions |
| Questionnaire distribution methods | Hand it out Put it online Send in post/email Ask people to collect it |
| Hand it out (Questionnaire distribution methods) | Target population gets questionnaire, but time-consuming |
| Put it online (Questionnaire distribution methods) | Data recorded//collected easily, but exclude those without internet access |
| Send in post/email (Questionnaire distribution methods) | wide-reaching, but unsure who is responding |
| Ask people to collect it (Questionnaire distribution methods) | easy to do, but people with strong views are more likely to get one then those without |
| Observation | Monitoring something. There are two ways to obtain the data counting, measuring |
| Counting (observation) | discrete data |
| Measuring (observation) | continuous data. However, this can never be exact and so a level of precision must be chosen and there will always be some error, which can vary by half the unit you measure either way |
| Reference | secondary sources of data |
| Census | Collects data about every member in population. Needs defined population. National census in UK every 10 years. Accurate, but hard on big populations, and if members are inaccessible. Also isn't an option if items are being used/damaged by investigation. |
| Data collection sheet | Where you record data. This should be designed carefully based on the type of data and how it will be analysed |
| Reference sources | Secondary sources of information. The source of reference data must be acknowledged. The reliability of the source must be considered - it could be biased, out of date, processed in the wrong format or incomplete |
| Reliability | the extent to which repeated measurements yield similar results |
| Validity | the extent to which a test measures what was intended |
| Sensitivity of the content matter | People may not truthfully answer a sensitive question, which could distort data and lead to incomplete responses. These are often about personal information |
| Random response solution to sensitivity of content matter | Random event make sure people who answer are anonymous + likely to answer truth. Eg, flip coin + heads pick yes + tails truthfully answer question. Then, estimate roughly half of yes; coin to get estimated answer. bigger sample more accurate results |
| Key features when planning data collection | Leading questions Avoiding biased sources Time factors Open questions Closed questions Opinion scales Interview Pilot questionnaire Pre-test (experiments) |
| Leading questions | where a question sways the response |
| Open questions | Allow any answers. This is good to know exactly what respondents think, but is harder to process |
| Closed questions | Have a fixed number of possible answers. They are easily processed and the answers aren’t ambiguous. But the answers are limited |
| Opinion scales | Give a statement and use a scale to gauge how strongly the respondent agrees or disagrees with it. |
| Interview | 1-1 conversations with each person in the sample. May complete a questionnaire's questions, or have a broader list of tips to talk about. Face-to-face, telephone or the internet |
| Interview Pros (compared to questionnaires) | More complex questions - can explain Higher response rate Know who filled it in Follow up answers |
| Interview Cons (compared to questionnaires) | Time-consuming Expensive May need to travel More likely to lie Could be recorded biasly (either on accident or deliberate) |
| Pilot questionnaire | Test on small group first. spot ambiguous, change hypothesis if an unexpected outcome, fix problems easier + cheaper+ estimate time+ costs of study + change distribution/ collection methods Make sure record made on problems + changes. over 1 is ok |
| Pre-test (experiments): | A trial-run of the experiment, to spot possible problems. Especially useful in cases where lots of subjects are used |
| Response rate | Number of responses divided by number questionnaires sent out |
| Non-responses | When people don’t respond. Could be for many reasons, but mostly they can't be bothered |
| Non-responses SOlutions | Clear questions with easy answers Follow up those who don't respond - impractical with a large/spread out sample Provide incentive Collect answers in person, but time-consuming and possibly impractical |
| Incomplete responses | when people don't answer every question. COuld be random, or for a particular question (e.g. a sensitive question) |
| Incomplete responses Solutions | Clear questions Be careful with sensitive questions |
| Incomplete responses How to deal with them | Discard incomplete questions - which may lead to bias results Follow up people to fill them in Replace it with an educated guess or a mean - but this can really affect the data Analyse incomplete responses to work out the causes |
| Possible problems with recorded data | Outliers Missing data values. Wrong format. Different units may have been used. |
| Outliers | Values that seem out of place with the rest of the data. They may be accurate but extreme, or incorrectly recorded. |
| Cleaning data | Fixing problems before processing - correct inaccuracies, deal w/ outliers, put data in format + remove units. Outliers may be removed, or analysis methods unaffected by extreme values, may lead to inaccuracy. Data re-recorded, but may be impractical |
| Extraneous variables | Variables (not explanatory or response) that affect experiment. Must be constant so don’t influence response variable. ensures change in response due to changes in explanatory. May not always be possible to control, so should be identified to be accurate |
| Control groups | Members of sample randomly assigned to two groups. experimental group treated but control group used as reference - giving certainty that data is valid. All extraneous variables should be kept constant, so the only difference is the explanatory variable |
| Matched pairs | Before sample members split into groups, paired with another member that is similar. One member from each is randomly assigned into each group.Helps reduce variability + increase reliability. Other extraneous kept constant. This increases validity overal |
Created by:
FlashCardFun!