Hard texts to read

[Aelia Jusa]

I hope you didn't type all that out just so we could read it. It was on online text, right?

[Blade]

Yea its part of an online curriculum do you know how long it would to type all that out it would take me at least an 1/2 hour to an hour. lol I just copied and pasted it into the reply box.

------------------
Life and death are mearly shadows of the complexity of humans

[Cloudbringer]

whoa! Blade, that's mighty complicated material! Whew! need my techie pals to translate for us!

CloudLearner

------------------


Raindancer of the Laughing Hyenas Clan
Storm-Queen
Raven's Cloud
StormCloud of the Black Knight: Heart Mind Soul Forever
"To sleep, perchance to dream..."

[Epona]

Oh that's a classic! I don't understand a word of it.
I am sure I can dig something confusing out to add to the thread (not computer related though) but it is at home and I am at work, so you'll have to wait a bit!

------------------

Time is an illusion. Lunchtime doubly so.

Epona of The Laughing Hyenas
Proud winner of the 'Most Useless Post 250 Has Ever Seen' Award 2001. "I'd just like to thank my friends and family, without whom none of this would have been possible..."

[Sazerac]

Here's an example of one, Blade: the kind of stuff I had to study and understand in Graduate Statistical Analysis.

(the weak-stomached may wish to look away...also the math-phobic...)

________________________________________________

TESTING SUBSETS OF REGRESSORS IN FIDAP MULTIVARIATE MULTIPLE REGRESSION

Introduction

We have completed the integration and testing of a new addition to FIDAP Multivariate Multiple Regression: tests on subsets of regressors of interest. This new functionality allows hypothesis testing of different parts of the model using the extra sum of squares principle. The versions of FIDAP multiple regression on hippo and alw have been upgraded to incorporate this addition. The addition affects the following:

FIDAP Single-Subject Analysis (Main Menu #3) options:
(15) (Multivariate) Multiple Regression
(16) (Multivariate) Multiple Regression, modified to allow input stimuli not time-locked to the TR

The following sections address the i) the theory and operational equations used to test a subset of the regressors and ii) a practical guide to using the new feature in the FIDAP multiple regression.

I. Theory and Operational Equations for Subset Hypothesis Testing

The test on a subset of independent variables allows one to reject the null hypothesis that none of the dependent variables are predicted by the subset, or group, of independent variables. To assess this hypothesis the fit of a full model is compared to the fit of a reduced model. First, the B matrix of regressor coefficients may be written in partitioned form:

B = [Br Bd]

Br denotes the subset of regression coefficients to be kept in the reduced model, while Bd denotes the subset of regression coefficients to be deleted if the you suppose that they are not significant predictors of the response. Keep in mind that the NULL hypothesis being tested is that:

Bd = 0, or that the deleted variables are not significant predictors of the response

The full model is:

Y = XB + E, where Y are the dependent variables, X is the design matrix, B are the regression coefficients and E is the error matrix.

And the reduced model is:

Y = XrBr + E, where Xr contains the columns of the design matrix corresponding to the reduced matrix of beta coefficients, Br.

The test statistic is the ratio of Wilk's Lambda for the overall regression test in the full model (Lambda_f) to Wilks' lambda for the overall regression test in the reduced model (Lambda_r) which is computed as:

|Y'Y-B'X'Y|*(|Y'Y-Br'Xr'Y|)^-1

In the versions of FIDAP multiple regression that you have been using previously the reduced design matrix, Xr, and the reduced regression coefficient matrix, Br, contained all the regressors minus the one whose effect of interest that you were assessing in the case where you were assessing only that regressor of interest. Similarly, the reduced matrices Xr and Br contained only the regressors of no interest in the case where you were assessing the effects accounted for by all the regressors on interest (i.e., expressed as the familiar "int" map).

The point here is that the reduced matrix Xr may be constructed so that the effects of ANY subset of regressors of interest may be assessed.

The changes to the FIDAP multiple regression code allow you to easily perform these tests.

II. Changes Made to FIDAP Multiple Regression for Regressor Subset Hypothesis Testing

The FIDAP multiple regression code was changed in such a way that your existing scripts will continue to work in un-modified form if you do not want to perform subset hypothesis testing. However, if you choose option 15 or 16 under the Single Subject Analysis menu you will be asked the following additional question:

"You have the option of specifiying subsets of your regressors of interest, and then testing their contribution to the overall model."

"Do you want to test the significance of regressor subsets ?"

If you answer "n" to this question then FIDAP multiple regression will run exactly as it has previously, generating the same B_hat and wilks_lambda images as before.

If you answer "y" then you will be asked a series of new questions. The first will be:

"How many subsets of regressors or interest do you want to assess?"

You must type in an integer number here. There is no limit on the number of different subsets you may create and test.

For each subset (n of n subsets) you will be asked the following series of questions:

"How many regressors of interest do you want in subset #n?"

You must type in an integer here. You must input a number between 1 and the total number of regressors of interest. If you type in 1 then you will be testing the contribution of a single regressor of interest to the overall model, something which is already done by default for each regressor of interest.

Then you will be asked to enter the "index number" of each regressor of interest to include in the group. "The index number" is simply the number of the regressor file you entered in the script previously. If you had 14 regressors, 6 of which were regressors of interest, and you wanted to include your first regressor of interest in a particular subset then you would type 9; if you wanted to include the last regressor of interest then you would type 14. You will be reminded at each stage of questioning what your index numbers are for your regressors of interest. Error checking is done on the fly to make sure you don't specify an invalid number.

Once you have specified the regressor membership for each subset the multiple regression analysis will begin. You will notice that a matrix will be displayed indicating the subset membership. This matrix will be N-by-M, where N is the number of subsets and M is the number of regressors of interest. If for example you have specified three subsets (in a design with 14 regressors, 6 of which are interest regressors), the first testing for the contribution of regressors 9 and 10, the second testing the contribution of 12 and 13 and the third testing the contribution of 13 and 14 the matrix will look like:

1 1 0 0 0 0
0 0 0 1 1 0
0 0 0 0 1 1

This matrix will be written out along with the output images and called:

Subset_Regressor_Matrix_File.txt

This file exists as a record of subset membership and should be saved to remind you later of what regressor groupings map to the grouped regressor statistical output.

The output images for such a design would be named:

wilks_lambda_set1_p1vH2vE326.img
wilks_lambda_set2_p1vH2vE326.img
wilks_lambda_set3_p1vH2vE326.img

where "set" preceeds the subset number which corresponds to the row of the Subset_Regressor_Matrix_File.txt file. The wilks_lambda maps for each regressor of interest, and for the overall model (i.e., "omni") and all interest (i.e., "int") regressors will be written out as usual.

_________________________________________________

OK, enough torture for one session!



[This message has been edited by Sazerac (edited 09-05-2001).]

[Blade]

I get part of it because i have gone through pre-calculcus but the most of it just makes my head spin. I'm sure my brother would get it though he has taken math all the way up through linear Algebra and statics and Dynamics. Which make my head spinn too.


------------------
Life and death are mearly shadows of the complexity of humans

[This message has been edited by Blade (edited 09-05-2001).]

is there a wayn to turn html of?

[This message has been edited by Rikard (edited 09-05-2001).]

[Melusine]

LOL....whoa Rik, your post screwed up the entire thread, maybe you should edit! Anyway I agree Blade that it's difficult subject matter, and it's hard to understand if you're not into that kind of thing, but couldn't the same be said about any discipline?? I mean, gramatically your text is perfectly understandable, it's just the subject that will stump people. But so would you be if I gave you a text on a matter that you'd never studied.
What I find really hard to read are modern philosophical texts (thnk Gadamer, Lacen, Derrida). You read them twice and still you have no clue what they're saying, simply because they choose to put it so difficultly. However if you start realising what they're actually saying you'll notice it isn't so hard at all. I'm still suspecting philosophers of purposely writing so unintelligibly so us laymen won't realise philosophy isn't as hard as they make it out to be!!

------------------
Melusine, High Queen of Fluffies, Archbabe of the OHF, the LH,
the HADB and the SPAE(Society for the
Prevention of Acronym Extinction) &
Sultry Elflet

Your voice is ambrosia
Amy Brown Fantasy Art

[Blade]

You are probalby right Melusine and to Rickard's question there isn't a way that I know of to turn html (hyper text markup language i think is what it means) off becaue it is the language or one languages that the internet is writen in. Like reading a book that doen't have words in it, kinda usuless. Btw the 7 network layers thing was the second thing shown to us. The first was basic computer stuff like how different cables carried signals, and what parts of the computer did what.


------------------
Life and death are mearly shadows of the complexity of humans

[This message has been edited by Blade (edited 09-05-2001).]

[Sazerac]

Originally posted by Melusine: What I find really hard to read are modern philosophical texts (thnk Gadamer, Lacen, Derrida). You read them twice and still you have no clue what they're saying, simply because they choose to put it so difficultly. However if you start realising what they're actually saying you'll notice it isn't so hard at all. I'm still suspecting philosophers of purposely writing so unintelligibly so us laymen won't realise philosophy isn't as hard as they make it out to be!!

Melusine, agreed fully! I think the best philosophies are the ones so simply stated that a child can understand. There is a bit of "ivory-tower arrogance" that does go with certain academic studies of philosophy. Many profs write works that are readable only to a college-level graduate or above, because they (a) are unwilling to write in a manner that is intelligible for most of the population because of a degree of acadmeic arrogance, (b) they themselves don't fully understand the topic enough to be able to explain it simply, and so bury the parts of their ignorance in a flurry of obfuscation, or (c) they simply have forgotten how to write simply. A good reading of Strunk & White's "Elements of Style" would be recommended in the last case.

With few exceptions, most "ivory-tower" level philosophical treatises never reach popularity. The most timely ones are the ones that are reachable to all levels of humanity.



------------------