Qualifiers

[BG45]

Thanks for the prayers! First one is Friday...I think I'm done writing practice for Methods. Just study what I have now.

And I hadn't thought too much about Original Sin being linked to it! It works!

[BG45]

[center][b]Quantitative and Qualitative Executive Summary:[/b][/center]
[center][b][u]Quantitative Guide:[/u][/b][/center]

[b]General Layout[/b]: Intro, Sample, DV, IV, Survey, Limitations.

[b]Everything is a validity tradeoff. Bigger samp = better external. Smaller samp = better internal (better to fight history threats). When mailing, do selection letter, cover letter, and follow up. Units of analysis: People, groups, political and social orgs.[/b]

[b]Carmines & Zeller (1979):[/b]
[b][u]Validity[/u][/b][u]:[/u]
[u]Criterion: [/u]Agreement between instrument and criteria is measured with correlation. Limited in social sciences. Ex) Reported crime with a person’s criminal record.

[u]Content:[/u] How well does it represent what it claims? Review the lit, hard to put all in domain. Ex) Spelling test. Should include all words a student should know how to spell.

[u]Construct:[/u] Most used in social science. How well does the measure relate to other measures consistent with the theoretical hypothesis and construct being measured. Construct compared to something that theory and past research shows to be related to the topic of interest. 3 parts: Relate the theory and the construct and determine and explain the relationship (examine theory and previous lit). Test empirically. Explain the relationship between the theories and the variables. Ex) Collective efficacy and crime. Test with survey. Once tested, will explain in terms of current study. This method can be tested with factor analysis.

[u]Face Validity (DeVellis, 2003):[/u] What it sounds like.

[b][u]Reliability[/u]:[/b]

[u]Test-retest: [/u]Same group, 1-4 weeks, correlation, .70 or above according to Nunnaly (1979). Problems: external events such as maturation (unlikely in short time) or historical threats (event happened between testings).

[u]Internal consistency: [/u]Scale item reliability. Cronbach Alpha test correlates for same construct, .70 or above.

[b][u]Conceptualization and Operationalization Examples:[/u][/b]

[u]Ex) Violent Crime:[/u]
Conceptual: Can be defined as violations of laws involving a person’s safety.
Operational: Number of rapes, robberies, and assaults reported by the UCR for the year.
[u]Weight:[/u]
Conceptual: Relative mass of an object
Operational: Mass of object determined by place on scale.
[u]Religiosity:[/u]
Conceptual: Importance of religion in someone’s life.
Operational: Single item on importance of religion in life.
[u]Collective Efficacy:[/u]
Conceptual: Willingness of community members to stand up for the good of the community’s wellbeing.
Operational: Whether neighborhood involved in neighborhood watch and neighbor relationships.
[u]Punitive Beliefs:[/u]
Conceptual: Harsher punishment has created lower crime.
Operational: Items like incarceration, death penalty, etc. on scale of effectiveness to be rated

[b][u]Independent and Dependent Variable Examples: (Bold is IV, Italic is DV):[/u][/b]

Ex) [b]Collective efficacy[/b] go up, [i]violent crime[/i] goes down.

Ex) People with high [b]religiosity[/b] will more strongly [i]believe in punitive explanations[/i] for recent drop in crime.

Ex) People with low [b]self control[/b] are more likely to [i]commit crime[/i] than people with high [b]self control[/b].

[b][u]Sampling:[/u][/b]

[u]Simple Random: [/u]# list of population. Independent of each other. Use comp to select.

[u]Systematic Random[/u]: Randomly choose start, every Nth case.

[u]Stratified:[/u] Divide by logical segmentation, e.g. class rank.

[u]Multi-stage/Cluster:[/u] Divide and sub-divide. E.g. Cluster the schools in an area, then cluster students by rank, then take sample from those clusters of rank.

[b][u]Longitudinal Designs (Menard, 2002):[/u][/b]

[u]Longitudinal: [/u]Need two or more periods.

[u]Prospective Panel:[/u] 2 or more periods. Same variables and cases.

[u]Retrospective:[/u] Single period of collection, but for several periods worth of time.

[u]Repeated Cross-Sectional:[/u] Same variables, 2 or more times, non-identical cases.

[u]Needed for causality:[/u] Covariation, persisting relationship, measurable cause and effect. (Temp ordering needed for 3[sup]rd[/sup], first two can still be found in cross sectional).

[center][b][u]Qualitative Guide (Cresswell, 2007; Maxwell, 2005):[/u][/b][/center]

[u]Research Questions: [/u]Looks at meaning, context, process. Ex) How do x view reduction in crime rates? Is there a relate between X
and Y (coll eff and crime rates). What is relationship between bullying and having been the victim of bullying?

[u]Methods:[/u] Interview (1on1, focus groups, semi-structured, etc.), ethnographies, direct observation, participant obs., content analysis of records, etc. Always do memos/notes (body lang, what u think) helps with transcription and analysis.

[u]Analysis Strategy:[/u] Tape interviews, memo/notes, Qualrus or Nvivo.

[u]Gain access:[/u] formal/informal gatekeepers

[u]Sampling Strategies: Purposeful:[/u] For what want to learn. Use knowledge of group. [u]Snowball[/u]: Good way to find population you need.

[u]Human Subjects Protection:[/u] Inform consent, psch/emotional trauma issues, finance, legal, protected populations, etc.



[center][b][u]Afternoon Study Guide:[/u][/b][/center]

[u]R-squared:[/u] Coefficient of determination is R-squared. Beta = R in OLS regression, can square it. Also sum of squares regression/total sum of squares is R-squared. Amount of error explained by IV on DV. Can reduce error X% in predicting DV when taking IV into account.

[u]Unstandardized coefficient:[/u] is slope (b) on chart under unstandardized coefficient as IV (not constant). For this sample, 1 unit increase in (slope), is associated with (slope number) unit increase in DV.

[u]Standardized Coefficient:[/u] is Beta on chart. For this sample, on Average, 1 standard deviation unit increase in IV is associated with (Beta number) increase in DV. Beta is good for comparing multiple IVs and little else, because it standardizes measurements.

[u]95% Confidence Interval:[/u] Estimate the population slope (unstandardized coefficient) with 95% confidence interval. First need to meet the required assumptions of OLS regression.

[u]Assumptions:[/u]

[u]Specification Error:[/u] Include all relevant variables and exclude all irrelevant variables. If violated there is bias in coefficient.

[u]Measurement Error:[/u] States specifications for IV and DV, should be continuous and numeric. Bias in intercept and coefficient if violated.

[u]Independent Random Sampling:[/u] Sample randomly, everyone has an independent and random chance at being selected. Assume linear relationship. If violated, the results are skewed. Significant decrease in generalization.

[u]Error terms have a zero mean:[/u] Error cancels itself out, if violated this biases the coefficient.

[u]Error terms are not correlated with thr Independent Variables:[/u] What it says, bias if violated.

[u]Normal distribution of X and Y:Normal distribution of X and Y:[/u] X and Y are normally distributed. If not, can’t use CLT.

[u]Interval Level Data:[/u] You must have interval level data (or higher at ratio). If violate, you can’t run regression.

[u]Homoskedasticity: [/u]Equality of variance across all values of X. If violated, result in mis-specified model.

If assumptions are met, we can use [u]Central Limits Theorem (CLT[/u]): If we took the sample slopes and plot out, shape would be that of a normal/t-distribution depending on the degrees of freedom used. Shape of sample slope is equal to the population slope.

[u]Empirical Rule:[/u] Empirical distribution with known properties. If we have df we can find point where 95% of the cases are in the middle, with 2.5% to left and 2.5% to right. If 1 tailed hypothesis, where 95% are to the left and 5% are to the right. Our alpha is .05. DF = n-k. T-table says X value.; 95% of our slopes in sample distribution of slopes fall within x-value plus or minus standard deviations of the mean.

[u]Reverse Rule: [/u]Can reverse this rule, it reasons then we can put CI of x-value around all sample slopes. 95% will capture true population slope. 5% are Type One Error, claim to capture true population slope when we did not.

[u]Application to Sample: [/u]We apply to the sample with 1 slope. 5% chance of T1 error. Formula is: b plus or minus t (sb) CI is x____________y

[u]Interpretation:[/u] If lower of x-value – t(sb) is true, then on average, a 1 unit increase in IV is associated with x-value – t(sb) unit increase in DV. If upper of x-value – t(sb) is true, then on average, a 1 unit increase in IV is associated with x-value –t(sb) unit increase in DV. How wide? Similar conclusions? Is it useful? 5% T1 error risk.

[u]If they pull something else[/u]: Independent samples T test. Norm distribution of variables is needed. Groups near equal in size work. Indicate stat sig diff between groups.

[BG45]

That's 4,000 or so pages of Methods condensed. x.x Ditto with the Theory and Policy cards (or more pages on them).

I'm scared, but think I'm way more prepared this time.

[Amppax]

Prayers Buddy! This thread has been fascinating to read, at least the little bit I had. Its funny, I actually understand the math for the most part, but the rest has me completely baffled. And that is a first for me. Good Luck!

*Edit: by understand the math I mean i know what the different words mean. couldn't tell you how to do it Stats was fun.

Edited January 13, 2012 by Amppax

[BG45]

Thanks Amppax!! Glad it's been a good read, even if baffling at points. (it baffles me too at times.) And yeah...Stats is special. In undergrad I got a B. In my Masters I got a B. In both my doc classes on Stats I got Bs. I'm reliable apparently.

Methods done...my friend N reviewed it for me after I finished and turned it in, she passed the first time. I was iffy on my Dependent Variable, she is too. Otherwise she says it's a really good answer. Hoping for Pass or Low Pass...

Edited January 13, 2012 by BG45

[Amppax]

[quote name='BG45' timestamp='1311359734' post='2273909']
Thanks, though I know folks who would disagree with both counts there. [img]http://www.phatmass.com/phorum/public/style_emoticons/default/wink.gif[/img]

And Steubie!! If I end up at a FoP I should so call ahead!


Shover & Grabosky (2010) Topic: White Collar Crime and the Great Recession
1) Legislators believe that crime is a choice
[/quote]

If you're there for a FoP let me know! And L_D you in Pittsburgh?

[BG45]

Will do! I didn't go last semester, the new kids at the Catholic Student Association don't seem that big on going to possibly "charismatic stuff".

[CatherineM]

You're giving me Law School flashbacks.

[BG45]

Sorry Catherine, not pleasant I'm sure!

[BG45]

Today I went to Confession and we touched on Quals. Father blessed me afterwards and told me to go pass them like he knew I could.

[BG45]

Sorry to bump...I did pass! But I was reading something that made me think of this. When I worked under the head of our Psych department at my old school, he told me the story of one of his friends in his cohort who failed Quals and was found walking around naked the next day after snapping from the stress and muttering to himself.

Now it seems there is a possible, and I stretch only possible (as the university refuses to comment if he passed or not) link between failing Qualifiers and the shooting in Colorado at the Dark Knight Rises.
[quote]"Holmes had recently withdrawn from a competitive graduate program in neuroscience at the University of Colorado Denver, where he was one of six students at the school to get National Institutes of Health grant money. He recently took an intense three-part, oral exam that marks the end of the first year of the four-year program there, but university officials would not say if he passed, citing privacy concerns. The university said Holmes gave no reason for his withdrawal, a decision he made in June."[/quote]
[b] [/b]http://www.npr.org/templates/story/story.php?storyId=157166683

[Luigi]

CONGRATULATIONS!!!! on passing.

I remember your online study cards from last year, but I didn't know where you stood.

Good work!

[BG45]

Thanks Luigi! It's sad though, to think that his Qualifiers could have been a part of things. Plus that the people in the cohort behind me are all reading that story and realizing just how much stress can do to you.

[Chrysophylax]

[quote name='BG45' timestamp='1343055937' post='2458105']
Sorry to bump...I did pass! But I was reading something that made me think of this. When I worked under the head of our Psych department at my old school, he told me the story of one of his friends in his cohort who failed Quals and was found walking around naked the next day after snapping from the stress and muttering to himself.

Now it seems there is a possible, and I stretch only possible (as the university refuses to comment if he passed or not) link between failing Qualifiers and the shooting in Colorado at the Dark Knight Rises.

[url="http://www.npr.org/templates/story/story.php?storyId=157166683"]http://www.npr.org/t...oryId=157166683[/url]
[/quote]

Whoa, that is super interesting! and kinda creepy. I'm glad you passed so you didn't go insane!

[img]http://www.quizz.biz/uploads/quizz/108941/7_LyZl1.jpg[/img]
[b][size=5]"Stress; It's a killer, Sir."[/size][/b]

[she_who_is_not]

HaHa! My sister came home from the bar the other night and was telling me about her friend whose boyfriend developed schizophrenia in law school and she said, "I understand why you are the way you are now." I was like, "Um, thanks for understanding?" We had a guy go nuts at my law school. It was actually pretty exciting and switched the gossip away from who was sleeping with whom for about a week.