Experimental Group Designs Group Designs Simple Group Designs

  • Slides: 47
Download presentation
Experimental Group Designs

Experimental Group Designs

Group Designs • Simple Group Designs – one IV with 2 levels • 2

Group Designs • Simple Group Designs – one IV with 2 levels • 2 levels can be independent groups (Exp + Cont) • 2 levels can be repeated measurements (pre/post) • Complex Group Designs – one or more IVs • factorial designs – more than 2 levels on the IV – more than 1 DV • multivariate designs

Need to keep 2 things in mind simultaneously: • Independent Variable – # levels

Need to keep 2 things in mind simultaneously: • Independent Variable – # levels • 2 levels (bi-valent) --> SIMPLE • > 2 levels (multi-valent) --> COMPLEX – # variables • 1 IV (simple OR complex group design) • 2 or more IVs --> COMPLEX • Groups – Independent Groups – Repeated Measures

These 2 things can be mix-matched to come up with different design combinations. Ex:

These 2 things can be mix-matched to come up with different design combinations. Ex: 2 IVs with 2 levels each in an independent group design (2 x 2 independent group design)

Simple Complex Independent Groups Repeated Measures # Levels/# IVs 1 IV 2 levels (bivalent)

Simple Complex Independent Groups Repeated Measures # Levels/# IVs 1 IV 2 levels (bivalent) # Levels/# IVs > 1 IV > 2 levels (multivalent)

Simple Group Designs • Independent Group Designs – random selection designs – random assignment

Simple Group Designs • Independent Group Designs – random selection designs – random assignment designs – matched group designs – natural group designs • Repeated Measurement Designs • Simple Correlational Designs

Simple Group Designs • Involve 1 IV with 2 levels and 1 DV •

Simple Group Designs • Involve 1 IV with 2 levels and 1 DV • the levels of the IV can be independent groups or repeated measurements

4 Types of Independent Simple Group Designs • • random selection designs random assignment

4 Types of Independent Simple Group Designs • • random selection designs random assignment designs matched group designs natural group designs

Random Selection Designs • 2 groups are randomly selected from the same population •

Random Selection Designs • 2 groups are randomly selected from the same population • one group receives one level of the IV and the other group receives the other level • the effect of varying the IV is indicated by the difference between groups on the DV • this simple design doesn’t provide much control of subject variables such as age, gender, and education which researchers generally prefer to control

Random Assignment Designs • When only a small population of subjects is available, they

Random Assignment Designs • When only a small population of subjects is available, they can be randomly assigned to one group or the other. • This is the only difference from the random selection designs, that is, subjects are selected from a smaller population • subject variables are controlled by allowing them to vary randomly across both groups

Matched Group Designs • One or more variables that may affect the DV is

Matched Group Designs • One or more variables that may affect the DV is held constant between groups by matching the groups on those variables • Thus the problem of subject variability that was a problem in random selection designs is overcome with this design

Matched Group Designs • there are 2 types of matched group designs – groups

Matched Group Designs • there are 2 types of matched group designs – groups can be matched on the DV (e. g. , vocabulary skills, test scores, etc. ) – groups can be matched on variables that might affect the DV (e. g. , age, gender, education) • this design is more useful to CD researchers because of the small groups that are often available to researchers

Natural Group Designs • 2 groups selected from two different populations • In this

Natural Group Designs • 2 groups selected from two different populations • In this design, the IV is a difference between the groups created by nature that exists prior to the selection of the groups. • It is the effect of this IV (i. e. , difference between the groups) that is studied

Repeated Measurement Designs • This design has a single group of subjects in which

Repeated Measurement Designs • This design has a single group of subjects in which the two levels of the IV are varied within the same group of subjects • this design is used when there are not enough subjects available for two independent groups or when it is more efficient to carry out the experimental procedures within one group

Repeated Measurement Designs • The DV is assessed twice in a single group of

Repeated Measurement Designs • The DV is assessed twice in a single group of subjects • the difference between the two measurements demonstrates the effect of the IV • a problem with this design is the practice effect of repeating a measurement. • Another problem is the order effect of measurements administered to subjects. – To control for this, the researcher should use counterbalancing of the order of measurements to subjects.

Number of Subjects Needed for Simple Group Designs • 20 (10 per group for

Number of Subjects Needed for Simple Group Designs • 20 (10 per group for independent groups OR 20 for repeated measures) • In CDIS, the absolute minimum would be 10 subjects (5 per group for independent groups OR 10 for repeated measures)

Simple Correlational Designs • Two different measures are obtained from each subject in a

Simple Correlational Designs • Two different measures are obtained from each subject in a single group for determining if a relationship exists between the two measures – usually the IV and DV are not defined – it is difficult to interpret the relationships found in these designs

Complex Group Designs • Complex designs extend the simple group designs • more than

Complex Group Designs • Complex designs extend the simple group designs • more than 1 IV may be studied; more than 2 levels of the IV may be studied; and more than 1 DV may be examined • In addition, independent group designs and repeated measurement designs may be combined

Designs with more than 2 levels of the IV • Independent Group Designs: more

Designs with more than 2 levels of the IV • Independent Group Designs: more than two levels of the IV is examined – for example, comparing the effects of 3 levels of training (method a, method b, control) • Repeated Measurement Designs: assessing more than two things. – Order effects are still important so must counterbalance the order of presentation of tests, assessments, or measurements. – EX: IV - type of hearing aid; Levels - HA-1, HA-2, HA-3 same subjects are tested on all three levels (or HAs)

Designs with more than 1 IV (Factorial Designs) • Designs that vary two or

Designs with more than 1 IV (Factorial Designs) • Designs that vary two or more IV at same time can provide detailed information related to the complexity of the processes and disorders of communication • factorial designs can involve independent groups, repeated measures, or both (mixed factorial designs)

Designs with more than 1 IV (Factorial Designs) • The more complex the design,

Designs with more than 1 IV (Factorial Designs) • The more complex the design, the greater the number of experimental conditions (or cells) in the factorial design – two IV with 2 levels each is a 2 x 2 (4 cells) – three IV with 2 levels each is a 2 x 2 x 2 (8 cells) • with factorial designs, you can determine if there are main effects of each of the IVs as well as an interaction effect between the IVs

Designs with more than 1 IV (Factorial Designs) • Remember that factorial designs are

Designs with more than 1 IV (Factorial Designs) • Remember that factorial designs are COMPLEX DESIGNS • But, can have simple factorial designs and complex factorial designs

Simple Factorial Designs • The simplest factorial design has 2 IVs with 2 levels

Simple Factorial Designs • The simplest factorial design has 2 IVs with 2 levels each • the 2 IVs can be: – both independent groups – both be repeated measures – one independent group and one repeated measure (mixed factorial design)

2 x 2 Independent Group Design • Two groups that differ with respect to

2 x 2 Independent Group Design • Two groups that differ with respect to 2 different IVs, e. g. , – normal vs disordered; AND – male vs female

2 x 2 Independent Group Design Group Normal_____Disordered male Grp 1 Grp 3 Sex

2 x 2 Independent Group Design Group Normal_____Disordered male Grp 1 Grp 3 Sex female Grp 2 Grp 4

2 x 2 Independent Group Design | N |N | D |_D_____ M F

2 x 2 Independent Group Design | N |N | D |_D_____ M F no interaction

2 x 2 Independent Group Design | D |N | N |_D_____ M F

2 x 2 Independent Group Design | D |N | N |_D_____ M F interaction

2 x 2 Repeated Measurement Design • One group that received two different measurements,

2 x 2 Repeated Measurement Design • One group that received two different measurements, e. g. , tested HA 1 vs HA 2 in noisy vs quiet conditions – must control for order effects

2 x 2 Repeated Measurement Design Type of HA HA 1________HA 2 noisy Grp

2 x 2 Repeated Measurement Design Type of HA HA 1________HA 2 noisy Grp 1 condition quiet Grp 1

2 x 2 Repeated Measurement Design | 2 |2 | 1 |_1_____ N Q

2 x 2 Repeated Measurement Design | 2 |2 | 1 |_1_____ N Q no interaction

2 x 2 Repeated Measurement Design | 1 |2 | 2 |_1_____ N Q

2 x 2 Repeated Measurement Design | 1 |2 | 2 |_1_____ N Q interaction

2 x 2 Mixed Design • Two groups that receive some assessment, e. g.

2 x 2 Mixed Design • Two groups that receive some assessment, e. g. , normal vs disordered (independent group design) AND pretest vs posttest (repeated measure design)

2 x 2 Mixed Design Pretest Group Normal_______Disordered Grp 1 Grp 2 Posttest Grp

2 x 2 Mixed Design Pretest Group Normal_______Disordered Grp 1 Grp 2 Posttest Grp 1 Grp 2

2 x 2 Mixed Design Mild Treatment Tx A_______Tx B Grp 1 Severe Grp

2 x 2 Mixed Design Mild Treatment Tx A_______Tx B Grp 1 Severe Grp 2

Complex Factorial Designs • Factorial designs can be made more complex by increasing the

Complex Factorial Designs • Factorial designs can be made more complex by increasing the number of IVs, the number of levels of the IVs, or both • there can be 3 or more IVs and 3 or more levels of each IV • these designs are interpreted the same way as simple factorial designs, but there are many more possible outcomes

Complex Factorial Designs • The complex factorial design can provide more information about the

Complex Factorial Designs • The complex factorial design can provide more information about the complex interactions • the limitation of complex factorial designs are the number of subjects and the number of experimental conditions required by the design

Number of Subjects Needed for Complex Group Designs • 5 -10 subjects per independent

Number of Subjects Needed for Complex Group Designs • 5 -10 subjects per independent group or repeated measurement cell • Thus, a minimum of 20 -40 Ss would be needed for the 4 cells of a 2 x 2 factorial design and a minimum of 300 -600 Ss for the 60 cells of a 3 x 4 x 5 factorial design

Complex Correlation Designs • Simple correlation designs provide information about the relationship between 2

Complex Correlation Designs • Simple correlation designs provide information about the relationship between 2 variables • complex correlation designs provide more information about relationships • these designs are usually considered statistical techniques rather than designs

Types of Complex Correlation Designs • • • Partial correlation Multiple regression Factor analysis

Types of Complex Correlation Designs • • • Partial correlation Multiple regression Factor analysis Cluster analysis

Partial Correlation • Controls the effects that other variables may have on the relationship

Partial Correlation • Controls the effects that other variables may have on the relationship between 2 variables being examined. • Partial correlation adjusts the correlation between two variables that are being examined by eliminating the effects of their correlation with another variable

Complex Correlation Designs (con’t) • Multiple Correlation – Determines the relationship between criterion variables

Complex Correlation Designs (con’t) • Multiple Correlation – Determines the relationship between criterion variables and the predictor variable • Multiple Regression – Determines the relationship between each criterion variable and the predictor variable • Factor Analysis – Measures that are highly correlated with each other are grouped together with measures that are independent of each other • Cluster Analysis – a method for grouping subjects on the basis of patterns of deficit

Combined Correlation and Group Designs • Covariance designs – statistical procedures used to control

Combined Correlation and Group Designs • Covariance designs – statistical procedures used to control the effects of variables that might influence the IVs – similar in function to partial correlation • Multivariate designs – has more than 1 DV; it controls for the correlations between DVs

Combined Correlation and Group Design • Discriminant analysis designs – a statistical procedure used

Combined Correlation and Group Design • Discriminant analysis designs – a statistical procedure used to obtain a measure that will best differentiate two or more disordered groups from a normal group with regard to a number of variables on which the groups have been measured. – A weighted score is calculated for all the measures that best differentiate the groups – the proportion that each measure contributes to the total weighted score is varied until the weighted score that best differentiates the two groups is found.

Advantages of Group Designs • Isolate the effects of the IVs by systematically varying

Advantages of Group Designs • Isolate the effects of the IVs by systematically varying the levels of the IVs to determine the effects on the DVs – the IVs can be independent groups, repeated measures, or both • control the effects of other variables by allowing them to vary randomly in random selection designs, by holding them constant in matched group designs, or by systematically varying them in factorial designs

Advantages of Group Designs • Provide information about interaction effects of IVs on the

Advantages of Group Designs • Provide information about interaction effects of IVs on the DV • generalizability of findings • demonstrate causal relationships

Disadvantages of Group Designs • Difficulty obtaining large numbers of subjects – thus, groups

Disadvantages of Group Designs • Difficulty obtaining large numbers of subjects – thus, groups may be small and the number and levels of the IVs may be restricted – such restrictions limit the interpretations of results obtained with group designs and decrease the knowledge that can be obtained with these findings

Disadvantages of Group Designs • Group averages may not adequately represent the characteristics of

Disadvantages of Group Designs • Group averages may not adequately represent the characteristics of individuals • quantified measures of the DV may not provide enough information • may not apply to natural settings