Flashcards
Measurement Unit 10
Cards in this deck (42)
- 1Correlation between internalizing behavior and rule violations was not significant. The researcher concluded that externalizing behavior was associated with rule violations, and in contrast, that internalizing behavior was not associated with rule violations. What is the problem with this interpretation?lack of corr does not = discriminant validity
The problem is that scales tested with small samples will usually pass and scales tested with large samples will usually fail - 2In reporting statistical results, it is often important to report a descriptive statistic that indicates the size of an effect and an inferential statistic that tests the significance of the effect. In this case, what would the researcher report to describe the size of effect?r
- 3What inferential statistic would the researcher use to test for a significant difference?t
- 4To test for a significant difference (either by hand calculation or using R), the researcher will need to input four numerical values. These include three numerical values needed to calculate the inferential statistic, and one needed to determine the p-value for that statistic. What are these four values?r12, r13, r23, N
- 5To test for a significant difference between internalizing behavior and externalizing behavior in the magnitude of their correlations with rule violations, it is necessary to obtain numerical values for r12, r13, and r23. Using this example (which involves the variables internalizing behavior, externalizing behavior, and rule violations), define r12, then define r13, and then define r23.r12 = the correlation b/w rule violations and internalizing behavior scale
r13 = the correlation b/w rule violations and externalizing behavior scale
r23 = correlation between two scales validating (internalizing behavior and externalizing behavior) - 6Consider the following regression equation: ?hat = ?1?1 + ?2?2 + ?3?3. Assume you test this model using structural equation modeling (SEM). Draw a picture of the model that would be used to test this regression equation using SEM. (Note: on tests, you may simply be asked to recognize a correctly drawn model or to identify a mistake in a model that is drawn incorrectly.)
- 7If you test this model using SEM, what are the six parameters, not including variance components, that will be estimated?
- 8If you test this model using SEM, what are the four variance components that will be estimated?
- 9Using SEM, parameters for this model can be estimated from a variance covariance matrix. What is the size of this variance-covariance matrix (number of rows and columns)? How many unique covariances are in this matrix? How many variances are in this matrix?4 rows, 4 columns, 6 unique covariances, 4 variances.
- 10n structural equation modeling, models are typically estimated with no _____________ , as if _______________ , and this means that all parameters can be estimated from a _____________ . However, if a researcher wants to produce a model that has _____________, this can be done if the researcher includes _____________ in addition to using the ____________ to estimate model parameters.intercepts, all variables are centered, variance-covariance matrix. intercepts, means, variance covariance matrix.
- 11Consider the following regression equation: ?hat = ?1?1 + ?2?2 + ?3?3. What are three different types of constraints you could place on this equation if you were using structural equation modeling (SEM)? Using this equation, give an example for each type of constraint you list.Eliminate a path (B1=0) ,
Constrain path to equal a specific value (B1=.43),
Make two parameters equal (B1=B2) - 12Assume you wanted to estimate this model using least squares regression rather than SEM. With least squares regression. . .c. the five beta weights could be estimated using two equations.
- 13Assume you estimate this model and get the following parameter estimates: β1 = .2, β2 = .6 and β3 = .3. Based on these results (and assuming all variables are z-scores), fill in the values in the table below for the estimated variance-covariance matrix for these variables.
- 14In the previous question, you provided estimates for a variance-covariance matrix. If this were a real study, would these estimates be identical to or different from the actual variance-covariance matrix for the sample of participants in this study. State how you determined your answer.It is identical because
if use the model to re-create the
variance-covariance matrix it is perfect,
because you are using the betas as
they originally appeared in the data set, - 15In this example, what are the degrees of freedom for the model?0
- 16In this example, what is S – Σ? (give a numerical value)0
- 17Is this model saturated? Give an explanation for your answer using the words, “degrees of freedom.” Give another explanation for your answer using the words, “parameters” and “covariance.”Yes this is saturated because if you take the number of covariances (3) and subtract the number of parameters that you are estimating (3) the df = 0 , when you have 0 df the model is saturated
- 18Assume you estimate this model and get the following parameter estimates: β2 = .55 and β3 = .25. Based on these results (and assuming all variables are z-scores), fill in the values in the table below for the estimated variance-covariance matrix for these variables.
- 19In the previous question, you provided estimates for a variance-covariance matrix. If this were a real study, would these estimates identical to or different from the actual variance-covariance matrix for the sample of participants in this study. State how you determined your answer.Different: If you put constraints on your model it will not perfectly reproduce the variance covariance matrix, there will be an error in how it is recreated
- 20In this example, what are the degrees of freedom for the model?1
- 21In this example, what type of statement can you make about the nature of S – Σ?not equal to 0
- 22Is this model saturated? Give an explanation for your answer using the words, “degrees of freedom.” Give another explanation for your answer using the words, “parameters” and “covariance.Not saturated because you have fewer parameter estimates than elements in the variance covariance matrix, df = 1
- 23All methods used to define fit involveS - SIGMA
- 24The chi-square difference tests will test for a difference betweenS and SIGMA
- 25The chi-square difference test will test the null hypothesis thatmodel constraints = true population parameters
- 26When analyzing data using SEM, in what situation does a researcher want a model to have a good fit and in what situation does a researcher want a model to have a bad fit? In which of these situations would it be LEAST appropriate to use a chi-square difference test? Explain your answer. In which of these situations would it be most appropriate to use a chi-square difference test.. bad fit means constraints are not reasonable good fit provides good support for constraints,
Least appropriate chi square: when you have a hypothesized model that you are hoping to support, because your null hypotheses is that your model constraint = true population parameters
Most appropriate chi square: if your goal is to reject your null hypotheses - 27If a research wanted to use SEM to test for a difference between beta weights, what variable transformations need to be made before the analysis is conducted?standardize variables
- 28A researcher tested a regression equation with two predictor variables and obtained the following results: beta for x1 = –3.4, and beta for x2 = 8.2. If a researcher wanted to use SEM to test for a difference in magnitude of effect between the betas for x1 and x2, what variable transformations need to be made first?transform into z scores, reverse score scale 1 so betas have same sign
the two scales should be standardized, estimate unconstrained regression model (0 df and saturated, replica of matrix), estimate model with betas constrained to be equal (df=1, constraint, use chi sq to test for sig dif if sig constraint bad fit, and you have sig dif b/w betas)(sig chi square = discriminate validity) - 29The following questions come from information covered in the project. Assume you are using lavaan and you specify the following model: y ~ x1 + x2. a. Write the regression equation that is specified by this command..y = b1x1 + b2x2 + e
- 30The following questions come from information covered in the project. Assume you are using lavaan and you specify the following model: y ~ x1 + x2. Write a command you could use that would constrain the betas x1 and x2 to be equal.y ~ A*X_1 + A*X_2
- 31The following questions come from information covered in the project. Assume you are using lavaan and you specify the following model: y ~ x1 + x2.y ~ 3*X_1 +X_2
- 32The “Minimum Function Test statist is listed as “36.887.” What is this measuring?This measures the Chi-square value, which tests to see if the reproduction of the variance covariance matrix is something significantly different from the actual variance-covariance matrix.
- 33The degrees of freedom is listed as “1.” What does this mean?there was one constraint placed on the model
- 34The P-value (chi-square) is listed as “0.000.” What does this mean?this means that the constraint placed on the model significantly hampered its fit
- 35The estimator is listed as “ML.” What does this stand for?“ML,” which confirms that maximum likelihood was used to obtain parameter estimates.
- 36The “Minimum Function Test Statistic” is listed as “0.000.” What does this mean? Why is this value zero?The minimum function test statistic is a chi-square value testing the extent to which the reconstructed variance-covariance matrix (based on the parameters estimates from the model) is different from the actual, observed sample variance-covariance matrix. Because I did not place any constraints on the model (and it had zero degrees of freedom), the reconstructed variance-covariance matrix was identical to the sample variance-covariance matrix, the fit was perfect, and the chi-square value was zero.
- 37The degrees of freedom is listed as “0.” What does this mean?. the models was “saturated,” and this confirms the fact that I did not place any constraints on the model that could hamper its fit.
- 38In the last row, there is a value of “.0412” listed under the “Estimate” column for “.zCAHPS_YKS.” What is this value?This value is a residual (or unexplained) variance for the outcome variable. It is essentially the same thing as a mean-squared-error, and if you take the square root of this value, it should be quite close to the standard error of estimate from the least-squares regression analysis.
- 39Draw a picture of the model that was tested in this example (using boxes and arrows). Note: on tests, you may simply be asked to recognize a correctly drawn model or to identify a mistake in a model that is drawn incorrectly.
- 40What are the three parameters, not including variance components, that are included in your picture of this model? Of those three parameters in your picture, which are printed in the output above and which are not?B1 and B2 in output, B3 not in output
- 41What are the three variance components that were estimated in this model? Of these three variance estimates, which are printed in the output above and which are not?sd^2 ally, sd^2 confuse, sd_est not in equation
- 42The parameters for this model were estimated from a variance-covariance matrix. What is the size of this variance-covariance matrix (number of rows and columns)? How many unique covariances are in this matrix? How many variances are in this matrix?3 rows 3 columns, 3 variances, 3 covariances
Information
- Description
- graduate measurement Flashcards on Measurement Unit 10, created by abbi gutierrez on 29/03/2021.
- Created by
- abbi gutierrez
- Views
- 5
- Rating