spssforum.com

Hi,

I am doing a linear regression and among the independent variables are 5 factors. I want to validate my linear regression model on a testing/holdout sample, how do I calculate a factor score for the individuals in the holdout sample? Do I use the factor loadings and then how?

unclear what you have done, and what you want to confirm exactly.

Seems like you are taling about factor analysis and / or regression.

What do you want to confirm, the exact numbers, just whether you see the same pattern in another sample or...?
Can you please elaborate?

I am doing a logistic regression (not linear like I said before) and there was a problem with multicollinearity among almost all of the IVs except the demographics. PCA resulted in 10 factors and of these 10, 3 ended in my final logistic regression model among a couple of demographic variables. This was my training sample. Then I want to test this logistic model and how well it predicts in my testing sample, so I use the demographics and 3 factors but I do not know how I calculate a factor score for the new testing sample based on the factors that I got in the training sample. I think I use a formula which includes the set of variables/items in each factor but I am not sure how to calculate the scores precisely.

I hope this clarifies this better.

not exactly.

so you have x items and combined those into factors? or what do you mean by factors?

Yes, I did a principal component analysis and got 10 components/factors which each contain from 2 to 10 items/independent variables - in order to reduce the number of independt variables and also to deal with the multicollinearity...then I saved the factor scores for each case, then I used these component scores in the logistic regression. I want to make the component scores for a new data set to validate my model which contains some of the 10 components but I am unsure how to calculate the component scores for the new data set.

I don't really know how to clarify this better.

It is clear this time, before you did not mention you performed a factor analysis.

Look at these slides, I think they will be very inforamtive:
http://www.google.nl/url?sa=t&rct=j&q=w ... MQ&cad=rja

in addition, you might want to consider not using the component scores, but instead calculating factor scores yourself based on the loadings. Create a scale, so to say.

I see, sorry for that. I did a Principal component analysis (PCA).

Thank you very much for the link to the slides, they were helpful. But do you know if it is better to use all loadings of a component to calculate a factor score in the new data set, or use only loadings above a cut-off point, say 0.3. There are so many items (> 40) in the analysis and it would be time consuming to use all loadings...

Thank you again.

ps. in the slides you pointed at they mention split-sample validation of the factor analysis and for that they use Random number seed before they split the sample by creating a new split variable - can you please tell me what is the purpose of the Random number seed in this case and how do you choose a number for the seed??

about the random number seed: it gives a starting point for the randomisation, I think in this sheet they refer to a given number in their assignment, that is not applicable here.
you can also choose a random starting point by selecting the other option.

about the factors:
To be honest, I have only ever used factor analysis to create scales. For instance, in one situatio I created a questionnaire consisting of about 40 items, and I wanted to create scales (i had an idea of what these scales should be theoretically of course).

I used the rule of thumb that items should load >.40 on one factor and <.30 on all other factors.
A couple of items did not meet this criterium, therefore I did not include them at all.

So suppose you have 4 items that have a loading of >.40 on factor 1, and <.30 on all the other factors.
Then you can just take the mean of those 4 items (using compute), and use the new scale you just created in all other analyses.
I think that's actually the most common way to use factor analysis. So you don't use the "save" option from spss, because then every item is used for every factor.

OK, I think I get it now, thank you very much!

The cheap quinceanera dresses might be discovered in shops and via online shops for private use. All dresses are commonly tailor created for consumers who ask the dresses with their incredibly personal specifications. cheap quinceanera dresses might be worn together with many kinds of accessories, which consist of shoes and mind dresses.

There are many people, who are totally impressed by the results, which body shapers brought in their lives. These compression garments have made the lives of wearers easier and confident than ever before. if you are thinking that how could undergarments make the lives of people easy, then you should know that it is because of these compression garments, due to which women have got relieved, as they don't have to work hard inside gyms and go on strict diets, in order to reduce their weight and maintain their figure.

The problem with the dress is that Kimberly thinks it looks like a costume. The bust has a different fabric than what Kimberly wanted; she wanted it to be see-through. Oooh, classy. Fashion in the 1600s started becoming a little bit freer, and the western look came in style. Men started wearing leather boots, and leather cowboy hats, but the hose was still in style. Men linen shirts had deep cuffs.

It is a fantastic product for building different leg muscle cells, allows reduce pressure on feet and joints, and even allows. Whether its while travelling, working in the office or chilling with friends, we Cheap A-line wedding dress see people wired into their iPods or mobiles enjoying music. Your favourite music should only be enjoyed through a g
Bridal Dresses Sale

cocktail white dress page copy need to support each other as far as keyphrase use goes

_________________
short bridal dresses

You have cried and showed the ring to family and friends. Your parents/friends have thrown you this awesome engagement party. What could be better than what you are feeling at th. If you are conscious of the budget for your wedding ceremony, you have options in cheap wedding dresses. Among the most common weighing factors in the price charged by a photographer are location, time of year, experience and the time booked. Keep all of the factors in mind when you are looking,

One of the major reasons why some women dread attending parties is because they do not know what to wear or they have nothing stunning to wear. You may have even experienced this yourself, but you must not feel so hopeless because you can do something about it. There are several party dresses (party kjoler in dutch) to consider, but you have to make sure that you take some factors into consideration so that you will find the most stylish and stunning party dress that will complement you,

The skirt hangs to between the knees and the ankles, giving your body room to breathe but without going too short. To help your body look its best, the designers have draped the bodice. They have even placed an enticing little brooch at the waist to add a dash of sparkle and excitement to the dress,

Suni imediat. Oferi, din prima, cu toat膬 hot膬rrea, pre牛ul cerut chiar dac膬 este prea mare. La艧i telefon 艧i adres膬 艧i g膬se艧ti un motiv rezonabil s膬 amni prima 卯ntlnire cu dou膬, trei zile. Business clothes and evening dress comply with European norms of etiquette. For example, businessmen wear suits and ties and they often remove jackets during meetings. Businesswomen wear conservative suits in blue or gray, dresses, pantsuits, blouses and skirts
Classic Dresses

short bridal dresses as we'll see

_________________
short bridal dresses