Research Exam 2 pt. 2

p. 28-

higher levels of evidence depend on and are built up from

lower levels of evidence

the level of evidence of a study is affected by:

quality controls; a peer-reviewed empirical study which does not report quality controls should automatically be dropped 1-2 levels on the hierarchy, though typically not below a level 4

most studies you read will be where on the rating of study quality?

suggestive or inconclusive due to low quality controls

Mancopes et al. and Brian Weeks study application

• study is not a 2a bc it lacks some quality controls; it is probably a 2b or 3

• for relevance to Dr. Weeks, it probably has an inconclusive rating of study quality

• this is due to weak quality controls, not relevant to Weeks’ spinal muscular atrophy, and despite a strong design in the reviewed study, it is not clear EMST would work for Dr. Weeks

population

• in statistics, refers to all possible members of a group/category under study

• may be very broad, such as all people who have ever and will ever live, or narrow

• it is impractical to collect experimental data from this due to time, resource restictions, etc.

sample

• members of a population who participate in a research study

• as close as possible to random: equally likely to come and be assigned

• random of this allows us to assume members are similar to and representative of the characteristics of the population

• True randomness is impractical to obtain bc of the geographic, financial, temporal constraints or member preferences

Convenience sampling method

readily available but clearly non-random in one or more aspects; often include local residents or college students

Stratified sample method

people are assigned to subgroups, then randomly assigned to treatment groups; facotrs such as age and sex are balanced across treatment groups

Cluster random sampling method

people are first assigned to groups, then randomly assigned to treatment conditions; predefined groups such as medical centers, classrooms, or communities may be used for this type of method

dangers of sampling

over or under sampling

Increase sample size if: (pops and samples notes)

Decrease sample size if:

The DV is what we measure, but it can be measured…

in many ways

quantitative data measurement

• experimental research requires statistical analysis; there are 4 levels

• nominal

• ordinal

• interval

• ratio

what kinds of data are collected in questionnaires?

nominal and ordinal

nominal data

• categorical but unordered; form of name, attribute, label, etc.

• often data obtained by a survey: sex, ethnicity, marital status, etc.

• may include treatment outcomes: successful or not, transcription data

• categories are often treated as factors in statistical analysis, not DVs

ordinal data

• can be assigned rank, such as least to most

• categorical, such as grade in school, rather than continuous (such as age in seconds)

• differences btwn these categories may be uncertain or unknown; example, a score out of 3 on correct features rather than right/wrong

interval data

• interval: how much of an attribute or trait is present

• no true zero exists

• differences are known and can be manipulated by math

ratio data

• classic example is time

• has an absolute true zero

• equal intervals between values

• ability to perform all four basic arithmetic ops

descriptive statistics

• operations in a data set

• helpful for condensing large amounts of data

• specific to the sample

• common types include: measures of central tendency, measures of dispersion, z-scores, normal distribution

measures of central tendency

• identify the center of a distribution of scores

• mean, median, mode

• should ideally be the same; this occurs in a normal distribution

measures of dispersion

• variance: how much variation there is in a given set of scores

• standard deviation: most common measure of dispersion and most useful; simply the square root of the variance; difference between data points and mean on average

• range: highest score minus lowest score; always larger than standard deviation; crude, less specific

z-scores

• count the difference between an individual score and the mean in terms of the standard deviation

• convenient way to compare scores from different groups using different numerical scales

• also handy for normal distribution

normal/unimodal distribution

• essential part of inferential stats

• this is a bell-shaped, symmetric curve

• 68% +- 1SD of the mean

• 95% +- 2SD

• 99.7% +- 3SD

• examples: rainfall, language ability, response times

• mean, median, and mode are always the same

• type of shape varies based on SD and mode (more values in center = smaller SD, leads to a tall skinny graph)

sample size

• total number of overall participants who participated and were analyzed

• denoted as N

• subgroup denoted as n

standard deviation mental shortcut for data that is fairly symmetric

1. find middle value (approximate mean)

2. find a mini range of approximately 2/3 of the data in the middle

3. half of that width is approximately one SD

less accurate but alternative mental SD shortcut

calculate the range and divide by 4

confidence bands and [what] are similar concepts?

standard deviations

uniform/flat distribution type

• not very common in real life

• example where it occurs: powerball due to its random selections

bimodal distribution

• two values are the most frequent

• these are common in real life

• explained by differences; for example, a height/weight graph could be explained by gender difference