Show Menu
Cheatography
  1. Home >
  2. Cheat Sheets >

CS4249 Cheat Sheet by

Scales of Measur­ement

Scale
Descri­ption
Examples
Nominal
Caterg­orical; order doesn't matter
Gender: 1 (male), 2 (female)
Ordinal
Ordered values. Order matters, but not difference between values
Agreement: 1 (SD), 2 (D), 3 (Neutral), 4 (A), 5 (SA). Pain Scale (1-10)
Interval
Numeric. Difference between values is meaningful
Relative Temper­ature: °C, °F, pH
Ratio
Numeric. Zero and ratios are meaningful
Height, Weight, Absolute Temper­ature (K)
Measur­ement is the process of observing and recording the observ­ations collected as a part of a research effort.

Step 1: Define Research Questions

eg. How does your techni­que...
• Compare with altern­ative techni­ques?
Techniques
• For which target popula­tion?
Target users
• For what tasks?
Tasks
• In terms of what measures?
Perfor­mance measures
• In what context?
Other factors
Target users: need to be specific - students who have been using the desired medium consis­tently, for example
Perfor­mance measures: like speed, accuracy
Other factors: other than different techni­ques, what factors can influence the measures?

Step 2: Define Variables

IV
• Factors manipu­lated in the experiment
• Have multiple levels
DV
• Factors being measured
Control variables
• Attributes fixed throughout the experiment
• Confou­nders - attributes that vary and aren't accounted for
Random variables
• Attributes that are randomly sampled
• Increases genera­lis­ability
Confou­nders rather than IVs could have caused changes in DV.
They make it diffic­ult­/im­pos­sible to draw conclu­sions.
Order of presen­tation and prior experience are two important confou­nders that we need to control. (by counte­r-b­ala­ncing and proper sampling)

Step 3: Arranging Conditions (Withi­n-S­ubj­ects)

List the IV and their levels
eg.
Technique (2 levels: Gesture, Marking)
Menu depth (2 levels: 1, 2)
Determine counte­r-b­ala­ncing strategies for each IV
• Full counte­r-b­ala­ncing (n! condit­ions)
• Latin Square (n condit­ions)
• No counte­r-b­ala­ncing (seque­ntial) (1 condition)
Determine minimum no. of partic­ipants
Multiply all conditions together
Determine factorial arrang­ement of conditions
Put the permut­ations together
Determine arrang­ement for each partic­ipant
Condition reduction strategies:
• Pick the most import­ant­/in­ter­esting factors to test
• Run a few IVs at a time - if strong effect, include IV in future studies, otherwise, pick fixed control value for it

One-way ANOVA

Basic Idea: ANOVA tries to find the sources of this variance:
• due to difference between groups
• Variab­ility within each group
Total Variab­ility = Betwee­nGroup + Within­Group
𝑆𝑆𝑇=𝑆𝑆𝑀+𝑆𝑆𝑅
Ratio of Variab­ility
F = (𝑆𝑆𝑀/DFB) / (𝑆𝑆𝑅/DFW)
If the experiment is successful, then 𝑆𝑆𝑀>𝑆𝑆𝑅.
Betwee­n-group variab­ility will explain more variance than within­-group.
The bigger the F value, the smaller the p value, and the less like the null hypothesis (no differ­ence) is true.
Steps:
1. Calculate 𝑆𝑆𝑇
𝑆𝑆𝑇=𝑠_𝑔𝑟𝑎𝑛𝑑2 (𝑁−1)
DFT = (N-1)
2. Calculate 𝑆𝑆𝑀
𝑆𝑆𝑀=∑_i𝑛_𝑖 (𝑥 ̅_𝑖­−𝑥 ­̅_𝑔𝑟𝑎𝑛𝑑 )2
• sum of n difference of means from the grand mean
DFM = (No. of groups - 1)
3. Calculate 𝑆𝑆𝑅
𝑆𝑆𝑅=∑_𝑖𝑠_𝑖2 (𝑛_𝑖−1)
• sum of varianceno. of results in each group
DFR = total no. of results - no. of groups
Double check: 𝑆𝑆𝑇=𝑆𝑆𝑀+𝑆𝑆𝑅 & DFT = DFM - DFR
4. Calculate Mean Squared Error
MSM = SSM/DFM
MSR = SSR/DFR
5. Calculate F-ratio
F = MSM / MSR
if F is lower than value in F-table, then p < 0.05 results are statis­tically signif­icant

Behaviour Theories

Health Belief Model
Perceived Benefits v Perceived Barriers, Perceived Theat, Self-E­ffi­cacy, Cues to Action all contribute to Likelihood of Engaging in Health­-Pr­omoting Behaviour
Theory of Reasoned Action
Self-b­elief + Influenced beliefs, Attitudes, Intention Behaviour
Self-D­ete­rmi­nation Theory
Intrinsic (self-­ben­efit) v Extrinsic motivation (external benefits)
Goal Setting Theory
Basic idea: goal serves as a motivator, work harder as long as they believe goal is achiev­able. Importance in Clarity, Challenge and Feedback
Social Cognitive Theory
Cognitive, Enviro­nmental and Behavi­oural factors determine human behaviour
Fogg Behavi­oural Model
Behaviour = Motiva­tors, Ability, Triggers
• Motiva­tors: Sensation, Antici­pation, Social Cohesion
• Ability: Train or Simplify
• Triggers: Spark, Signal or Facili­tator
 

Testable Research Questions

Weak questions are untestable and broad
Stronger questions are more testable, but less genera­lizable

Step 4: Define Trials

Estimate the time for each trial
around 5-10 seconds?
Estimate the time for each condition
Time for each trial no. of trials for each condition
Balance the trials (so experiment is within 45 min)
Combine with the condition arrang­ement
Essent­ially, find the total time the experiment will take
Trials: a single repetition of a single condition
Typically want to have at least 3 trials per condition to increase reliab­ility
Consider time: trials should last for 45 minutes (excluding pre and post interv­iews)

Intera­ction Effect

Which t-test or ANOVA?

Cognition Processes

Attention
Perception
Memory
Learning
Reading, speaking & listening
Proble­m-s­olving, planning, reasoning & decisi­on-­making

Attention

Selecting things to concen­trate on at a point in time from the mass of stimuli around us
Focus on inform­ation that's relevant to what we are doing
Involves audio/­visual senses
Design implic­ations:
• Make inform­ation salient if it needs attending to
• make things stand out
•avoid cluttering interface

Perception

How inform­ation is acquired from the world, and transf­ormed into experi­ences
Design repres­ent­ations that are readily percei­vable
Implic­ation:
• Group inform­ation
• Text should bne legible and distin­gui­shable from the background

Memory

Stages of memory:
• Encoding
• Storage
• Retrieval
Encoding:
• Determines which info is attended to in enviro­nment + how it's intepreted
• Context affects extent to which info can be retrieved - different context difficult to recall
Implic­ations:
• Focus attent­ion/no compli­cated procedures
• Recogn­ition over recall
• Provide various ways of encoding and retrieving info (searching v history)
Storage:
Sensory Memory:
• shorte­st-term memory, acts like a buffer for stimuli retrieved
• Ability to remember and process info at same time
• Inform­ation will decay within 10-15s
• Extended by rehearsal, hindered by interf­erence
Long-term Memory:
•Decla­rative Memory (factual info):
 • Semantic Memory (general) + Episodic Memory (personal knowledge)
• Procedural Memory (skill­s/h­abits)
Retrieval:
• Intern­al/­Ext­ernal stimuli for retrieval cues
• Encoded at same time as memory

Cognitive System Principles

Uncert­ainty Principle
𝑇=𝐼𝐶𝐻
where T = Decision time, H = log2(n+1) (where n is the no. of choices)
Variable Rate Principle
More effort Faster processing (ie. cycle time
Cycle time also diminishes with practice: 𝑇_𝑛=𝑇_1 𝑛−𝛼
Fitts' Law
𝑇_𝑀=𝑎+𝑏 log_2⁡­(𝐴/𝑊+1)
where A = distance to target, W = error tolerance

Trans-­the­ore­tical Models

5 Stages of Change
Pre-co­nte­mpl­ation, Contem­pla­tion, Prepar­ation, Action, Mainte­nance
Processes of Change
Consci­ousness raising, Social libera­tion, Goal setting, Helping relati­ons­hips, Rewards
Processes of change can be applied to 5 stages of change.
Each person will value different processes differ­ently.
  

Statistics

We use sample statistics to estimate/make inferences about population parameters
Due to uncert­ainty and variab­ility, conclu­sions and estimates may not always be correct.
Need measures of reliab­ility
• Confidence interval
• the confidence that the true population value of a parameter falls within a confidence interval
• affected by: variation & sample size
• Level of signif­icance
•“P value”, α
• the prob. of rejecting the null hypothesis when it is actually true (Type I error)
• ie. concluding that there is a difference when there may be no actual difference
• signifies the probab­ility that the difference is due to chance
Level of Signif­icance Thresholds
• Not signif­icant (p>.1; p=n.s.)
• Marginally signif­icant (p<0.1)
• (Fairly) signif­icant (p<.05)
• (Good) signif­icant (p<.01)
• (Excel­lently) signif­icant (p<.001)

Some Formulae

Cumulative Percentage

Central Limit Theorem

As the sample size gets larger...
The mean of sample means approaches the population mean
The standard error of the sameple means = the standard deviation of the population mean
𝑆𝐸=𝑠/√𝑛=√(𝑉𝑎𝑟𝑖𝑎𝑛𝑐𝑒/𝑛)

SED Between 2 Samples

2-Sample t-test

Small sample sizes not normal distri­bution
Use t-dist­rib­ution
Steps:
1. Calculate mean difference
2. Calculate SD
3. Calculate no. of SDs away from 0
4. Calculate df = smaller n - 1
5. Calculate p-value, for signif­icance (which p-value is it closest to)
If given desired confidence interval, steps:
1. Given desired CI
2. Get no. of SDs away from 0 from t-table
3. Calculate margin of error in units ((2) SD)
Difference between groups more likely to be signif­icant if:
• Large difference between means
• Small SD or large n in each group
Assump­tions:
• Continuous variable
• Indepe­ndent samples
Also called the indepe­nde­nt-­samples t-test
Other tests:
• One-sample t-test (sample v constant)
• Paired­-sa­mpled t-test (withi­n-s­ubj­ects, repeated measures)
• One-way ANOVA

Cognitive Heuristics

Affects
where emotions influence decisions
Availa­bility
where people overes­timate the importance of inform­ation available to them
Confir­mation Bias
where we only listen to inform­ation that confirms out precon­cep­tions
Halo Effect
where an outcome in one area is due to factors from another
Framing Effect
where the words used push listeners in a certain direction
Implic­ations: watch out for biasing your partic­ipants.

Structural Equation Modeling

Design Strategies for Lifestyle Behaviour Change

Abstract & Reflective
Unobtr­usive
Public
Aesthetic
Positive
Contro­llable
Trendi­ng/­His­torical
Compre­hensive

Nielsen Heuristics

Visibility of system status
Match system and real world
User control and freedom
Consis­tency and standards
Error prevention
Recogn­ition over recall
Flexib­ility and efficiency of use
Aesthetic and minimalist design
Help users recognise, diagnose, recover from errors
Help and docume­ntation
 

Metadata

  • Languages:
  • Published: 25th November, 2018

Comments

No comments yet. Add yours below!

Add a Comment

Your Comment

Please enter your name.

    Please enter your email address

      Please enter your Comment.