Experimental Data Analysis Sheet



k-Group Non-Experimental Data Analysis Sheet

Later in this lab you will analyze data from a controlled experiment. It will be pretty simple – having the data for the two analyses nicely organized into a data set. However, most of the data analyses behavioral scientists do are different from that upcoming analysis in three ways: 1) most of the data we analyze are non-experimental 2) most of the data we analyze are imbedded within much larger data sets, and 3) most of the data we analyze require some sort of “data preparation” before we can analyze it. That’s what we are going to practice now -- working with large non-experimental data sets.

BG Analysis – Walk-Through ( Does the number of siblings relate to how much you like people?

The interpersonal dataset includes the following variable.

16 -- Liking People Scale (LPS) -- single scale score

This 15-item instrument measures one aspect of interpersonal orientation, the general liking of other people. Interpersonal orientation plays a significant role in one’s social development and adjustment. The theoretical point of departure of the LPS is that the degree of liking of people influences whether one approaches or avoids social interaction. The instrument has utility, then, for monitoring intervention in cases of social isolation, shyness, and antisocial behavior.

The researcher wanted to know if people from 2-parent families with different numbers of siblings differed as to how much they like people. Specifically, the hypothesis was that those with siblings will like people more than those without siblings. To do this we must select the correct participants from the sample, construct any necessary variables, and then perform the proper statistical analysis.

Step 1: Following the handout, select only those respondents from 2-parent families (family = 2) from the dataset for analysis

Step 2: Make a new variable called numsibs ( starting with “siblings” and recode it so that:

No siblings is coded ( 1

1 sibling is coded ( 2

2 or more siblings is coded ( 3

Step 3: Perform a BG ANOVA to see if there are mean Liking People differences for the different sibling groups

Omnibus BG ANOVA

| |No siblings |1 sibling |2+ siblings |

|Mean | | | |

|Std | | | |

F = df = , MSerror = p = N = k = n =

LSD, Pairwise Comparisons, Effect Sizes & Statistical Decision Errors

Do we need to perform LSD pairwise comparisons to test the RH? Why or why not?

Do the LSD analysis anyway ( LSDmmd =

| Pairwise comparison ( |No vs. 1 sibling |No vs. 2+ sibling |1 vs. 2+ siblings |

|Mean diference ( | | | |

|LSD result ( | | | |

|Type of Stat Error risked ( | | | |

|Pairwise effect size ( | | | |

|Power Problem? ( | | | |

RH: Testing

RH1: Those with no siblings will have lower average LPS score than those who have 1 sibling

Is this RH: fully, partially or not supported? Explain your answer.

RH2: Those with no siblings will have lower average LPS score than those who have 2+ sibling

Is this RH: fully, partially or not supported? Explain your answer

RH3: Those with 1 sibling will have the same average LPS score as those who have 2+ siblings

Is this RH: fully, partially or not supported? Explain your answer

Overall, is this set of RH: fully, partially or not supported?

Following the example in the SPSS handout, write-up the results of the omnibus-F and the pairwise comparisons. Provide a table or a figure for the data (using SPSS or doing it in word, etc. your choice) and properly refer to that table or figure in the write-up.

BG Analysis – On You Own ( Does a student’s plans following graduate related to how much they work while in school?

The personal success dataset includes data about undergraduate student’s plans following graduation and how many hours they currently work each week.

The researcher wanted to know if these variables are related for people who’s family income when they were seniors was between $50k – 120k. Specifically, they expect to find that those who plan to go to professional school will work the least hours while in college, and those who plan to go to graduate school will work the most hours while in college.

Step 1: Following the handout, select only those respondents who’s family income was greater than 50000 and less than 120000 (the if statement in the select cases window should say “faminc > 50000 and faminc < 120000”

Step 2: Perform a BG ANOVA to see if there are mean “number of hours of work each week” differences for students with the different after college plans.

Omnibus BG ANOVA

| |Find Work |Graduate School |Professional School |

|Mean | | | |

|Std | | | |

F = df = , MSerror = p = N = k = n =

LSD, Pairwise Comparisons, Effect Sizes & Statistical Decision Errors

Do we need to perform LSD pairwise comparisons to test the RH? Why or why not?

LSDmmd =

|Pairwise comparison ( |Work vs, Grad |Work vs. Prof |Grad vs. Prof |

|Mean diference ( | | | |

|LSD result ( | | | |

|Type of Stat Error risked ( | | | |

|Pairwise effect size ( | | | |

|Power Problem? ( | | | |

RH: Testing

RH1: Those who plan to go to work will work more hours in college than those who plan to go to professional school.

Is this RH: fully, partially or not supported? Explain your answer.

RH2: Those who plan to go to graduate school work more hours in college than those who plan to go to professional school.

Is this RH: fully, partially or not supported? Explain your answer

RH3: Those who plan to go to work will work fewer hours in college than those who plan to go to graduate school.

Is this RH: fully, partially or not supported? Explain your answer

Overall, is this set of RH: fully, partially or not supported?

Following the example in the SPSS handout, write-up the results of the omnibus-F and the pairwise comparisons. Provide a table or a figure for the data (using SPSS or doing it in word, etc. your choice) and properly refer to that table or figure in the write-up.

.

WG Analysis Walk-through ( comparing aspects of Interpersonal Dependency

The interpersonal data set some of you used for your project includes the following set of subscales…

25-27 -- Interpersonal Dependency Inventory (IDI) - Emotional Reliance on Others, Lack of Self-confidence, Assertion of Autonomy subscales

The IDI is a 48-item instrument designed to measure the thoughts, behaviors, and feelings revolving around need to associate closely with valued people. The theoretical base for the IDI is a blend of psychoanalytic, social learning, and attachment theories emphasizing the importance of excess dependency to a range of emotional and behavioral disorders. Based on an initial pool of 98 items, the 48-item scale was developed using factor analysis.

The researcher wanted to know if women undergraduates differed across these three aspects of the construct. The specific hypothesis was that Emotional Reliance on Others would have the highest mean and Lack of Self-confidence will have the lowest mean.

Step 1: Following the handout, select only the females from the dataset for analysis

Step 2: The different scales have different numbers of items, so comparing their means wouldn’t make sense (the subscales with more items will probably have the larger means). So, we have to compute new versions of each subscale, by dividing each subscale score in the dataset by the number of items in that subscale.

Compute a new variable called relper ( dividing rel by 18 (the number of items in that subscale)

Compute a new variable called lackscper ( dividing lacksc by 17

Compute a new variable called autoper ( dividing auto by 14

Step 3: Perform a WG ANOVA using the three new variables

Omnibus WG ANOVA

| |Emotional Reliance |Lack of Self-confidence |Assertion of Autonomy |

|Mean | | | |

|Std | | | |

F = df = , MSerror = p = N = k = n =

LSD, Pairwise Comparisons, Effect Sizes & Statistical Decision Errors

Do we need to perform LSD pairwise comparisons to test the RH? Why or why not?

LSDmmd = .

|Pairwise comparison ( |ER vs. LoSC |ER vs. AoA |LoSC vs. AoA |

|Mean diference ( | | | |

|LSD result ( | | | |

|Type of Stat Error risked ( | | | |

|Pairwise effect size ( | | | |

|Power Problem? ( | | | |

RH: Testing

RH1: Participants will have higher Assertion of Autonomy than Lack of Self Confidence.

Is this RH: fully, partially or not supported? Explain your answer.

RH2: Participants will have lower Emotional Reliance than Assertion of Autonomy

Is this RH: fully, partially or not supported? Explain your answer

RH3: Participants will have lower Emotional Reliance than Lack of Self Confidence

Is this RH: fully, partially or not supported? Explain your answer

Overall, is this set of RH: fully, partially or not supported?

Write-Up

Following the example in the SPSS handout, write-up the results of the omnibus-F and the pairwise comparisons. Provide a table or a figure for the data (using SPSS or doing it in word, etc. your choice) and properly refer to that table or figure in the write-up.

WG Analysis On Your Own ( comparing aspects of Compulsiveness

The self-description data set some of you used for your project includes the following set of subscales…

21. Compulsiveness Inventory (CI) – Indecision and Double-Checking, Order and Regularity, and Detail and Perfection subscales

The CI is an 11-item scale designed to measure behaviors that are common in the “normal’ population(non-pathological compulsiveness).

The researcher wanted to know if only children differed in their “expression” of compulsiveness across these three dimensions. Specifically, based on the existing literature, the hypothesis was that score would be much higher on the obsessive-compulsive scale than on either the indecision-doublechecking scale or the order-regularity scale, which would be equivalent to each other.

Step 1: Following the handout, select only the only children (siblings = 0) from the dataset for analysis

Step 2: The different scales have different numbers of items, so comparing their means wouldn’t make sense (the subscales with more items will probably have the larger means). So, we have to compute new versions of each subscale, by dividing each subscale score in the dataset by the number of items in that subscale

Compute a new variable called dbper ( dividing ind_dblck by 5 (the number of items in that subscale)

Compute a new variable called ordregper ( dividing ord_reg by 4

Compute a new variable called detperper ( det_perf by 2

Step 3: Perform a WG ANOVA using the three new variables

Omnibus WG ANOVA

| |Indecision & Double-checking |Order & Regularity |Detail & Perfection |

|Mean | | | |

|Std | | | |

F = df = , MSerror = p = N = k = n =

LSD, Pairwise Comparisons, Effect Sizes & Statistical Decision Errors

Do we need to perform LSD pairwise comparisons to test the RH? Why or why not?

LSDmmd =

|Pairwise comparison ( |IDC vs. O&R |IDC vs. DP |O&R vs. DP |

|Mean diference ( | | | |

|LSD result ( | | | |

|Type of Stat Error risked ( | | | |

|Pairwise effect size ( | | | |

|Power Problem? ( | | | |

RH1: Participants will have higher Detail & Perfection scores than Indecision-Double checking scores

Is this RH: fully, partially or not supported? Explain your answer.

RH2: Participants will have higher Detail & Perfection scores than Order & Regularity scores

Is this RH: fully, partially or not supported? Explain your answer

RH3: Participants will have equivalent Indecision & Double checking and Order & Regularity scores

Is this RH: fully, partially or not supported? Explain your answer

Overall, is this set of RH: fully, partially or not supported?

Write-Up

Following the example in the SPSS handout, write-up the results of the omnibus-F and the pairwise comparisons. Provide a table or a figure for the data (using SPSS or doing it in word, etc. your choice) and properly refer to that table or figure in the write-up.

X² Analysis – Walk-Through ( Gender and Internal/External Locus of Control

The interpersonal dataset includes gender and a measure of Internal-External Locus of control:

Internal Versus External Locus of Control Scale -- single scale score (higher scores mean more external attribution)

An internal locus of control indicates that an individual believes that he or she is responsible for the reinforcements experienced; in effect, that the person’s actions, characteristics, qualities, etc. are prominent determinants of the experiences being queried. An external locus of control, however, indicates that the person views his or her outcomes as being primarily determined by external forces, whether they be luck, social context, other persons, or whatever.

Reference: J.B. Rotter Generalized expectancies for internal versus external control of reinforcement. Psychological Monographs, 80, 1966 (whole No. 609).

The researcher wanted to know if there is a relationship between gender and locus of control. The RH: was that the % of females would be higher for those with External locus of control than for either those with Internal locus of control or those with Undifferentiated locus of control

Step 1: We have to convert the current quantitative Locus of Control measure to a categorical variable. Those with a locus score less than 8 (i.e., 7 or less) will be coded 1 and identified as “Internal’; those with a locus score greater than 13 (i.e., 14 and higher) will be coded 3 and identified as “External”; those with a score between 8 and 13 (i.e., those with scores larger than 7 and less than 14) will be coded 2 and identified as “Undifferentiated”.

Step 2: Perform a X² using the two qualitative variables

Omnibus X²

| |Internal-External Locus of Control |

|Gender |Internal |Undifferentiated |External |

|Male | | | |

|Female | | | |

X² = df = p = N = k = n =

Pairwise X²-crtical, Pairwise Comparisons, Effect Sizes & Statistical Decision Errors

Do we need to perform pairwise X² comparisons to test the RH? Why or why not?

X²-critical =

| Pairwise comparison ( |Int vs. Undif |Int vs. Ext |Undif vs. Ext |

|% Female |vs. |vs. |vs. |

|X² result ( | | | |

|Type of Stat Error risked ( | | | |

|Pairwise effect size ( | | | |

|Power Problem? ( | | | |

RH: Testing

RH1: The % of Females is larger for those with External locus of control than for those with Internal locus of control.

Is this RH: fully, partially or not supported? Explain your answer.

RH2: The % of Females is larger for those with External locus of control than for those with Undifferentiated locus of control.

Is this RH: fully, partially or not supported? Explain your answer

RH3: The % Females is the same for those with Internal locus of control and those with Undifferentiated locus of control.

Is this RH: fully, partially or not supported? Explain your answer

Overall, is this set of RH: fully, partially or not supported?

Following the example in the SPSS handout, write-up the results of the omnibus-F and the pairwise comparisons. Provide a table or a figure for the data (using SPSS or doing it in word, etc. your choice) and properly refer to that table or figure in the write-up.

X² Analysis – On Your Own ( Family Type & Greek Organization Membership

The self-description dataset includes the type of family a person grew up in and whether or not they are in a Greek organization while in college.

The researcher wanted to know if there is a relationship between family type and Greek organization membership. The RH: was that the % of Greek Members would be higher for those raised in single-parent families and 2-parent families than it would be for those raised in step-parent families. Further, the researcher expected that the % of Greek members would be the same for those raised in single- and 2-parent families.

Step 1: The fourth Family Type category in the dataset, Foster Parents, occurred very few times. Including this category in the analysis could produce unreliable results, both because of population representation issues and statistical power limits. So, we need to select those participants in the data set who have Family Type variable values of 1, 2 or 3 (excluding those with a value of 4).

Step 2: Perform a X² using the two qualitative variables

Omnibus X²

| |Type of Family |

|Member of |1-parent |2-parent |Step-parent |

|Frat/Soro? | | | |

|Independent | | | |

|Memberf | | | |

X² = df = p = N = k = n =

Pairwise X²-crtical, Pairwise Comparisons, Effect Sizes & Statistical Decision Errors

Do we need to perform pairwise X² comparisons to test the RH? Why or why not?

X²-critical =

| Pairwise comparison ( |1-parent vs. 2-parent |1-parent vs. step-parent |2-parent vs. step-parent |

|% Greek |vs. |vs. |vs. |

|X² result ( | | | |

|Type of Stat Error risked ( | | | |

|Pairwise effect size ( | | | |

|Power Problem? ( | | | |

RH: Testing

RH1: The % of Greek members is larger for those with single-parent families than for those in step-parent families.

Is this RH: fully, partially or not supported? Explain your answer.

RH2: The % of Greek members is larger for those with two-parent families than for those in step-parent families.

Is this RH: fully, partially or not supported? Explain your answer

RH3: The % of Greek members is the same for those with two-parent families than for those in one-parent families.

Is this RH: fully, partially or not supported? Explain your answer

Overall, is this set of RH: fully, partially or not supported?

Following the example in the SPSS handout, write-up the results of the omnibus-F and the pairwise comparisons. Provide a table or a figure for the data (using SPSS or doing it in word, etc. your choice) and properly refer to that table or figure in the write-up.

-----------------------

Grading

Non-Experimental

Walk-throughs ______ 5

BG On Your Own ______ 20

WG On Your Own ______ 20

X² On Your Own ______ 20

Total Graded Points ______ - ______ (points lost - why?)

Assignment grade out of 65 points __________

................
................

In order to avoid copyright disputes, this page is only a partial summary.

Google Online Preview   Download