Introduction to Latent Class & Latent Profile Analysis
- Shared screen with speaker view

Garth

57:35

Unable to open the pdf slides password does not seem to work

Nyssa Snow-Hill

58:12

I had to download the zip file.

KAITLYN STORMES

58:17

@garth I think you have to download the zip file. I noticed you canâ€™t click on them directly.

peishanyen

01:27:56

Could we use the result LCA as the golden standard of the clinical diagnosis for depression groups (no depression, mild depression, major depression), instead of using threshold for total score? Because some cases with middle total scores does not mean they do not have major depression. So, LCA is a better way to identify TRUE categorial outcome?

Jin Park

01:38:06

I have a question.

Jin Park

01:39:56

What is the yardstick (cutpoint) to say that an item belongs to a the group? Is it intuitively understood or is there are particular number or above it should be categorized in the group?

peishanyen

01:43:10

Totally understood! THANK YOU!!

Jin Park

01:43:50

either way.

Mary Hannan

01:44:21

Other than making sure the survey items used ask questions that can clearly delineate how the groups are qualitatively different, are there any other "requirements" for the survey that would be used as your indicators? (i.e. does that survey have to have a certain number of survey questions, certain threshold of reliability/validity, etc.) And as the researcher, if we are using survey items that are continuous, can we as the researcher categorize them to turn them into categories (if it has been done in the literature)?

Jin Park

01:47:43

Thank you.

Patricia

01:54:49

Would it be helpful to use NPP, PPP, sensitivity or specificity data if available to select items from a larger pool of items?

Mary Hannan

01:58:37

Yes, thank you!

Garth

02:01:11

Could you share that paper or citation?

Patricia

02:02:41

Thank you.

Jin Park

02:04:11

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2726788/

Patricia

02:10:41

Class 1 - Non-drinker

Noah Tye

02:10:46

Never drink before

Vikki Lynn Rompala

02:11:05

no alcohol use

Jordan Smith

02:11:31

Heavy drinker

Rachel Lane

02:11:58

Current drinker

Larisa A Burke

02:12:11

Binge drinker

Rebecca

02:12:46

Just quit drinking

Jen Glayzer

02:12:54

occasional drinker

Noah Tye

02:13:09

Had a drink in the year

Garth

02:18:16

reflective

Garth

02:18:21

Yay I won!

Noah Tye

02:18:24

Can you give a definition?

Jen Glayzer

02:19:30

boxes

Patricia

02:19:36

the dots?

Uchechi Mitchell

02:19:40

the circle

Mary Hannan

02:19:41

inside the circle

copefeurer

02:19:48

in the latent drinking classes variable

Noah Tye

02:40:44

Is there a guide book (how to manual) for how to run the examples in SAS? I have never used SAS before.

Patricia

02:41:02

And in SPSS?

Sabrina Katherine Young

02:47:01

Are we going to talk at all about how to use these probabilities as a dependent variable or weight in other regressions?

Sabrina Katherine Young

02:49:49

yes

Larisa A Burke

02:49:52

yes

Larisa A Burke

02:50:38

1?

Garth

02:50:40

infinity?

Mary Hannan

02:50:41

1

Jen Glayzer

02:50:43

1

Oksana Pugach

02:50:47

1

Oksana Pugach

02:51:01

-1

Naoko Muramatsu

02:51:56

0

Jordan Smith

02:51:58

0

copefeurer

02:51:58

0

Jen Glayzer

02:52:34

1

Ayo Olagoke

02:52:36

1

Larisa A Burke

02:52:37

1

Uchechi Mitchell

02:52:37

1?

Mary Hannan

02:52:41

1

Rachel Lane

02:52:52

1 or 0?

Oksana Pugach

02:53:25

0.5

Uchechi Mitchell

02:53:26

0.5

Jordan Smith

02:53:27

.5

Patricia

02:53:30

.5

Jen Glayzer

02:53:33

1/number of classes

Judith Schlaeger

02:53:34

.5

Jen Glayzer

02:57:46

high vs low achievers

Mark B Lockwood

02:58:42

users vs. nonusers

Vikki Lynn Rompala

02:58:47

low users vs. high users

copefeurer

02:59:13

polysubstance users, legal substance users

Garth

03:03:34

High homogeneity and high separation sound like the same thing to me (?)

Garth

03:03:52

Got it-nevermind

Patricia

03:48:37

Can these analyses be run in SPSS?

Larisa A Burke

03:50:11

I'm getting an error proc LCA not found

copefeurer

05:30:58

is there an issue if your AIC or BIC are a negative value?

Jen Glayzer

05:33:14

Would you ever have to choose from a model that has a better AIC/BIC but worse hetero/homogeneity?

Daejun Park

05:37:48

I have two questions about entropy.First, entropy is traditionally considered to select a model. But then I wonder why it's not considered in this table.Second, Can I choose the model with the value of entropy under 0.8?If the five classes has a value of 0.7 entropy, while six classes has a value 0.8 entropy, then which model should I choose?Can I still choose the five-classes with under the 0.8 entropy value?

Jen Glayzer

05:40:41

thank you! very helpful

Daejun Park

05:43:12

Thank you very much

Larisa A Burke

05:51:38

I got an error message in SAS

Larisa A Burke

05:51:47

WARNING: The information matrix could not be inverted during standard error calculation.It was singular and/or not positive definite.This is sometimes due to either insufficient degrees of freedom ora boundary value estimate for a parameter.Valid standard errors could not be computed.We suggest trying to use RHO PRIOR statement.Please see the users' guide for details.

KAITLYN STORMES

05:52:20

I tried but am having trouble with the remote access at my institution for the software. Can you run the code and show us what the output looks like or how to briefly interpret it?

Lan

06:04:25

Class 3

Mark B Lockwood

06:05:48

Frequent user in favor of use

Jen Glayzer

06:06:26

class 4

Patricia

06:06:32

Class 4: Approver

Patricia

06:07:32

Class 2:

Jin Park

06:08:08

Class 1: curious experimenters

Larisa A Burke

06:11:07

Would you expect different salutations based on the program? Mplus estimates look different than SAS.

Garth

06:11:14

How does MPLUS do this?

Jin Park

06:12:46

I tried to run exercise 1 in mplus7 but there is a problem. Is it because I am using mplus ver. 7?

Garth

06:13:21

identification

Jin Park

06:14:10

Mplus VERSION 7.4 (Mac)*** ERRORThe input file does not contain valid commands:/Users/yangjinpark/exercise-1-mplus-input.inp

Garth

06:17:33

Your MPLUS codes has STARTS = 500 50; so this is for 7.4?

Jin Park

06:18:04

Got it. Thank you!

Garth

06:18:26

Thanks

Garth

06:31:16

Would this be a nominal logistic regression model of LC as dependent variable?

Jen Glayzer

06:35:25

how do decide if a variable is a grouping variable vs covariate? Does it have to do with the type of variable ie dichotomous vs continuous?

copefeurer

06:37:52

is it possible to extract group membership for each participant, and conduct post-hoc analyses in another program, such as SPSS?

Jen Glayzer

06:39:12

thank you!

Rachel Lane

06:39:23

You talked earlier about the uncertainty associated with assigning individuals to classes. Do these kinds of predictive models do anything special to account for that uncertainty?

Noah Tye

06:39:24

What would a negative beta mean?

Lan

06:40:56

Is the y axis probability?

Mark B Lockwood

06:41:46

Are these models used in AI algorithms for risk stratification?

Rachel Lane

06:41:49

Thank you

Mark B Lockwood

06:42:49

yes

Mark B Lockwood

06:45:19

excellent. Thank you!!

Lan

07:04:29

Is LPA another way than K-means approach for cluster analysis?

Lan

07:12:04

Thanks

Naoko Muramatsu

07:17:54

Can LCA/LPA allow correlated measurement error, as in SEM?

Lan

07:18:04

Just confirm: no LPA in SAS?

Daejun Park

07:18:20

It's about LCA. When I run LC models with covariates, can I import the text data saved in Mplus into other software like stata?I mean I wonder if I can use the 'C' variable which is created after running LCA in mplus and then use it into stata to run regressions?

Garth

07:18:50

In general, how would things change if you had a mix of categorical and continuous indicators?

Mark B Lockwood

07:19:09

Any chance for another workshop focusing on longitudinal data (LTA)? This was wonderful!

copefeurer

07:21:55

to confirm, is there any way to extract group membership for individual subjects, and then look at predictors of class membership, differences between the groups, or class membership as a predictor of outcomes in another statistical package, or this should only be done incorporating your predictor or outcome into the same model that is predicting class membership?

Daejun Park

07:24:25

Got it. Thanks!

Garth

07:24:52

Ahhh

copefeurer

07:25:08

Thank you! I also would have a lot of interest in LTA!

MUrio

07:25:40

How should we go about conducting a power analysis for LCA? What are the parameters we need to worry about? Is it different for LPA and LTA?

Maria Granros

07:25:46

Are there any general guidelines for sample size for conducting LPA/LCA?

Mark B Lockwood

07:26:03

great!

Larisa A Burke

07:28:17

In the mean time, I saw Bethany's has an LTA video is on you tube! https://www.youtube.com/watch?v=c9Y3drFgIKk&t=770s

copefeurer

07:29:31

it does, thank you!

Jang Ho Park

07:29:49

In BCH method, (I believe this is a method with the distal outcome), can we run BCH with covariate variable?

Garth

07:32:39

Could you replace logistic regression with probit regression in MPLUS for LCA, advantages/disadvantages?

MUrio

07:32:47

Thank you!

Garth

07:33:14

Could you restate which paper has the power analysis discussion (I spaced)

Tim Johnson

07:33:40

thank you, Bethany, for taking the time to give us this presentation. Very clean and very helpful.

MUrio

07:37:18

Listed in Agenda-Reading List:

MUrio

07:37:21

Power AnalysisDziak, J. J., Lanza, S. T., & Tan, X. (2014). Effect size, statistical power and sample size requirements for the bootstrap likelihood ratio test in latent class analysis. Structural Equation Modeling, 21, 534-552. doi: 10.1080/10705511.2014.919819 PMCID: PMC4196274Nylund, K. L., Asparouhov, T., & Muthen, B. O. (2007). Deciding on the number of classes in latent class analysis and growth mixture modeling: A Monte Carlo simulation study. Structural Equation Modeling, 14, 535-569. doi: 10.1080/10705510701575396

Maria Granros

07:38:49

Thanks!

Jang Ho Park

07:42:33

Thank you. Can I also find the syntax comparing models in Mplus web note 21?

Garth

07:42:38

Echo Tim, this was an excellent workshop, thanks so much for your time and effort with this!

Patricia

07:43:04

This was a great workshop. Thank you so much!

ashleykendall

07:43:06

Thank you so much, Bethany!!

Erin Marie Emerson

07:43:46

Thank you Bethany! Really great workshop and presentation!

copefeurer

07:43:49

This was fantastic! Thank you!

Larisa A Burke

07:43:50

Thanks so much!

Mary Hannan

07:43:52

Thank you! This was wonderful!

Lan

07:43:55

Thanks a lot!

Patricia

07:43:56

I hope we're on your Center's mailing list so we know about all the upcoming events.

Oksana Pugach

07:43:57

Thank you!

MUrio

07:43:58

Very helpful! Thank you!

Daejun Park

07:44:01

This was amazing. Thank you!

peishanyen

07:44:06

Thank you a lot!!!

Maria Granros

07:44:08

Very helpful! Thanks!!

Uchechi Mitchell

07:44:09

Thank you very much! This was great.

Linda Schiffer

07:44:09

Thank you!

Lillian

07:44:23

Thank you!! Very informative.