factor analysis for binary variables
Locked
factor analysis for binary variables
 
|
Hi everyone,
The 70 variables that will be subjected to factor analysis are binary(coded 0 and 1). May I know your thoughts regarding the methods to use in extracting the factors, as well as the rotation methods. Thank you in advance for your help. Eins --------------------------------- Get your preferred Email name! Now you can @ymail.com and @rocketmail.com. ===================== To manage your subscription to SPSSX-L, send a message to [hidden email] (not to SPSSX-L), with no body text except the command. To leave the list, send the command SIGNOFF SPSSX-L For a list of commands to manage subscriptions, send the command INFO REFCARD |
Re: factor analysis for binary variables
|
|
Eins,
Factor analysis does not lend itself well to binary data. You need to do Latent Class Analysis, or loglinear modeling. If you must do FA (which would not be recommended) then use vector coding for the binary variables instead of dummy coding 0 and 1. See Factor Analysis - practical Issues by Kim and Mueller (SAGE Publications); and Loglinear Models with Latent variables by Hagenaars (SAGE publications). Neda Faregh -----Original Message----- From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Eins Bernardo Sent: Saturday, October 11, 2008 9:22 PM To: [hidden email] Subject: factor analysis for binary variables Hi everyone, The 70 variables that will be subjected to factor analysis are binary(coded 0 and 1). May I know your thoughts regarding the methods to use in extracting the factors, as well as the rotation methods. Thank you in advance for your help. Eins --------------------------------- Get your preferred Email name! Now you can @ymail.com and @rocketmail.com. ===================== To manage your subscription to SPSSX-L, send a message to [hidden email] (not to SPSSX-L), with no body text except the command. To leave the list, send the command SIGNOFF SPSSX-L For a list of commands to manage subscriptions, send the command INFO REFCARD ===================== To manage your subscription to SPSSX-L, send a message to [hidden email] (not to SPSSX-L), with no body text except the command. To leave the list, send the command SIGNOFF SPSSX-L For a list of commands to manage subscriptions, send the command INFO REFCARD |
Re: factor analysis for binary variables
|
|
Actually, it is exploratory factor analysis(EFA). What are the violations if I use EFA for the binary variables? Why latent class analysis?
Eins Neda Faregh <[hidden email]> wrote: Eins, Factor analysis does not lend itself well to binary data. You need to do Latent Class Analysis, or loglinear modeling. If you must do FA (which would not be recommended) then use vector coding for the binary variables instead of dummy coding 0 and 1. See Factor Analysis - practical Issues by Kim and Mueller (SAGE Publications); and Loglinear Models with Latent variables by Hagenaars (SAGE publications). Neda Faregh -----Original Message----- From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Eins Bernardo Sent: Saturday, October 11, 2008 9:22 PM To: [hidden email] Subject: factor analysis for binary variables Hi everyone, The 70 variables that will be subjected to factor analysis are binary(coded 0 and 1). May I know your thoughts regarding the methods to use in extracting the factors, as well as the rotation methods. Thank you in advance for your help. Eins --------------------------------- Get your preferred Email name! Now you can @ymail.com and @rocketmail.com. ===================== To manage your subscription to SPSSX-L, send a message to [hidden email] (not to SPSSX-L), with no body text except the command. To leave the list, send the command SIGNOFF SPSSX-L For a list of commands to manage subscriptions, send the command INFO REFCARD --------------------------------- Get your new Email address! Grab the Email name you've always wanted before someone else does! ===================== To manage your subscription to SPSSX-L, send a message to [hidden email] (not to SPSSX-L), with no body text except the command. To leave the list, send the command SIGNOFF SPSSX-L For a list of commands to manage subscriptions, send the command INFO REFCARD |
Difference in sample size a in one-way ANOVA?
|
|
In reply to this post by Neda Faregh
Hi ALL,
Is extreme difference in sample sizes a problem in one-way ANOVA? I asked this because I am conducting one-way ANOVA for five groups with sample sizes of 34, 40, 53, 70 & 180. Thanks. Eins --------------------------------- Get your new Email address! Grab the Email name you've always wanted before someone else does! ===================== To manage your subscription to SPSSX-L, send a message to [hidden email] (not to SPSSX-L), with no body text except the command. To leave the list, send the command SIGNOFF SPSSX-L For a list of commands to manage subscriptions, send the command INFO REFCARD |
Re: factor analysis for binary variables
|
|
In reply to this post by E. Bernardo
Eins,
Factor Analysis, including Exploratory Factor Analysis, assumes linearity and normality of the underlying data. In FA, each variable is assumed to be a weighted sum of at least two underlying factors. FA analysis starts with a covariance matrix of the relationships between continuous variables (observed) measured at interval or ratio level. Factor analysis is a form of latent variable analysis, for continuous data. But when the observed variables are discrete, as in your case, you need a categorical data analogue to FA which is latent class analysis (LCA). Binary variables give you counts which can be analyzed using contingency tables and log linear modeling. And LCA uses a multivariate frequency table as its starting point. As far as I know SPSS does not have LCA capabilities. There are a number of other programs available including MPLUS by Muthen & Muthen. Neda Faregh -----Original Message----- From: Eins Bernardo [mailto:[hidden email]] Sent: Saturday, October 11, 2008 11:08 PM To: [hidden email]; [hidden email] Subject: RE: factor analysis for binary variables Actually, it is exploratory factor analysis(EFA). What are the violations if I use EFA for the binary variables? Why latent class analysis? Eins Neda Faregh <[hidden email]> wrote: Eins, Factor analysis does not lend itself well to binary data. You need to do Latent Class Analysis, or loglinear modeling. If you must do FA (which would not be recommended) then use vector coding for the binary variables instead of dummy coding 0 and 1. See Factor Analysis - practical Issues by Kim and Mueller (SAGE Publications); and Loglinear Models with Latent variables by Hagenaars (SAGE publications). Neda Faregh -----Original Message----- From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Eins Bernardo Sent: Saturday, October 11, 2008 9:22 PM To: [hidden email] Subject: factor analysis for binary variables Hi everyone, The 70 variables that will be subjected to factor analysis are binary(coded 0 and 1). May I know your thoughts regarding the methods to use in extracting the factors, as well as the rotation methods. Thank you in advance for your help. Eins --------------------------------- Get your preferred Email name! Now you can @ymail.com and @rocketmail.com. ===================== To manage your subscription to SPSSX-L, send a message to [hidden email] (not to SPSSX-L), with no body text except the command. To leave the list, send the command SIGNOFF SPSSX-L For a list of commands to manage subscriptions, send the command INFO REFCARD _____ Get <http://sg.rd.yahoo.com/ph/mail/domainchoice/mail/signature/*http:/mail.prom otions.yahoo.com/newdomains/ph/> your new Email address! Grab the Email name you've always wanted before someone else does! ====================To manage your subscription to SPSSX-L, send a message to [hidden email] (not to SPSSX-L), with no body text except the command. To leave the list, send the command SIGNOFF SPSSX-L For a list of commands to manage subscriptions, send the command INFO REFCARD |
Re: Difference in sample size a in one-way ANOVA?
|
In reply to this post by E. Bernardo
At 01:59 AM 10/12/2008, Eins Bernardo wrote:
> Is extreme difference in sample sizes a problem in one-way > ANOVA? I asked this because I am conducting one-way ANOVA for five > groups with sample sizes of 34, 40, 53, 70 & 180. I'm far from the leading ANOVA expert here, but I don't think you have a problem. First, ANOVA estimates and statistics inherently take cell size into account. Second, while you have large differences in cell size, I wouldn't call them 'extreme'. The ratio of your largest to your smallest cells is less than an order of magnitude, and I wouldn't start talking about 'extreme' until at least an order of magnitude difference. More cause for concern is the absolute size of the smallest cell; but I think 34 will do very nicely. There are other questions. Notably, do the samples in all the cells have approximately equal variances? Bad things can happen to your ANOVA if they don't; for what is likely to go wrong, and what to do about it, I refer you to others on the list. -Best of luck, Richard ===================== To manage your subscription to SPSSX-L, send a message to [hidden email] (not to SPSSX-L), with no body text except the command. To leave the list, send the command SIGNOFF SPSSX-L For a list of commands to manage subscriptions, send the command INFO REFCARD |
One-way ANOVA, disparate sample sizes, robustness, power.
|
|
If heterogeneity of variance is a problem, ask for a Welch Test -- One
Way ANOVA, Options, Welch. Yes, robustness may suffer with disparate sample sizes. Power will too, as it is a function of the harmonic sample size, which, with fixed total N, is larger the more nearly equal the sample sizes. Of course, if your effect is significant, that is of little concern (unless you get upset about confidence intervals that are wider than you would like). Cheers, Karl W. -----Original Message----- From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Richard Ristow Sent: Sunday, October 12, 2008 11:39 PM To: [hidden email] Subject: Re: Difference in sample size a in one-way ANOVA? At 01:59 AM 10/12/2008, Eins Bernardo wrote: > Is extreme difference in sample sizes a problem in one-way > ANOVA? I asked this because I am conducting one-way ANOVA for five > groups with sample sizes of 34, 40, 53, 70 & 180. I'm far from the leading ANOVA expert here, but I don't think you have a problem. First, ANOVA estimates and statistics inherently take cell size into account. Second, while you have large differences in cell size, I wouldn't call them 'extreme'. The ratio of your largest to your smallest cells is less than an order of magnitude, and I wouldn't start talking about 'extreme' until at least an order of magnitude difference. More cause for concern is the absolute size of the smallest cell; but I think 34 will do very nicely. There are other questions. Notably, do the samples in all the cells have approximately equal variances? Bad things can happen to your ANOVA if they don't; for what is likely to go wrong, and what to do about it, I refer you to others on the list. -Best of luck, Richard ===================== To manage your subscription to SPSSX-L, send a message to [hidden email] (not to SPSSX-L), with no body text except the command. To leave the list, send the command SIGNOFF SPSSX-L For a list of commands to manage subscriptions, send the command INFO REFCARD ===================== To manage your subscription to SPSSX-L, send a message to [hidden email] (not to SPSSX-L), with no body text except the command. To leave the list, send the command SIGNOFF SPSSX-L For a list of commands to manage subscriptions, send the command INFO REFCARD |
Re: factor analysis for binary variables
|
|
In reply to this post by Neda Faregh
Is it not that LCA is analogous to Cluster Analysis, not to factor analysis? I am not sure, but it is discussed here: http://ourworld.compuserve.com/homepages/jsuebersax/faq.htm#uses
Eins Neda Faregh <[hidden email]> wrote: Eins, Factor analysis does not lend itself well to binary data. You need to do Latent Class Analysis, or loglinear modeling. If you must do FA (which would not be recommended) then use vector coding for the binary variables instead of dummy coding 0 and 1. See Factor Analysis - practical Issues by Kim and Mueller (SAGE Publications); and Loglinear Models with Latent variables by Hagenaars (SAGE publications). Neda Faregh -----Original Message----- From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Eins Bernardo Sent: Saturday, October 11, 2008 9:22 PM To: [hidden email] Subject: factor analysis for binary variables Hi everyone, The 70 variables that will be subjected to factor analysis are binary(coded 0 and 1). May I know your thoughts regarding the methods to use in extracting the factors, as well as the rotation methods. Thank you in advance for your help. Eins --------------------------------- Get your preferred Email name! Now you can @ymail.com and @rocketmail.com. ===================== To manage your subscription to SPSSX-L, send a message to [hidden email] (not to SPSSX-L), with no body text except the command. To leave the list, send the command SIGNOFF SPSSX-L For a list of commands to manage subscriptions, send the command INFO REFCARD --------------------------------- Is Bangus or Tilapia more delicious? Tell us what you think in Yahoo! Answers! ===================== To manage your subscription to SPSSX-L, send a message to [hidden email] (not to SPSSX-L), with no body text except the command. To leave the list, send the command SIGNOFF SPSSX-L For a list of commands to manage subscriptions, send the command INFO REFCARD |
«
Return to SPSSX Discussion
|
1 view|%1 views
| Free forum by Nabble | Edit this page |