Simple Lineal Regression in RPL & other issues
in [HP48]
|
Prev: Newer ROMs for HP 50g?
Next: System Fonts
From: lacro on 14 Jun 2008 12:43 I want to make a very-very-simple application of this subject. I've noticed I became rusty with the usage of my current HP 50g. All this mess is due to the fact that I should get all the stat data, BUT WITHOUT THE SDAT matrix values, although I might only provide the 6 initial summary statistics as the local variables defined. Besides the much better suggestions for programming I'm sure I will receive and that they are welcome, I´m still even trying to find out my gross mistakes -let alone the right concepts-. My own next challenge would be to compile the Urroz' and many other alternatives to solve the statisitical distributions. There is stiil no one I could find which can I depend on. eg. DPROB library published at Eric's Site, besides using a unique and different nomenclature, lacks a lot of comings and goings. The best I knew from Stuart Sidders costs a small fortune, and the partial one from him (converted and ported by Gaak), lacks many distributions (LN, Gamma, Weibull, Pascal, t, Chi2 and F-St at least). On the other hand, I never found out how to use the LIN function in order to linearize expressions of the kind like 'Y = alfa * X^beta' to get 'ln Y=ln alfa + beta * ln X' Partial Mathematic Expectation (for Mean) is one subject which I understand (pls correct me if I'm mistaken) is not considered in the ongoing statistic courses like here in Argentina (thanks to Professors Mermoz and García from the Statiscal Courses from the School of Engineering of the University of Buenos Aires). Perhaps known by other name, I'm glad to share here their work in a PDF around that area in order to develop HP solutions. . http://www.fi.uba.ar/materias/7103/Esperanza%20matematica%20parcial%201a%20Parte.pdf http://www.fi.uba.ar/materias/7103/Esperanza%20matematica%20parcial%202a%20Parte.pdf http://www.fi.uba.ar/materias/7103/esperanzaparcial.pdf http://www.cuspide.com/isbn/9871104448 http://www.cuspide.com/isbn/9502312953 Another ancient pending issue, a reasonable 4x emulator for DOS for my HP 200LX. Either I need a ROM I don't know how to get, or several files ara missing. Both professors mentiones earlier, also developed a very accurate solution in this environment (which I would be glad to share for academic purposes). ----------------------------------------------- Back to my initial inquiry, where there are three dots it should figure a summation sign, where there is a block previous to TAG command it should figure the arrow sign "->", and I used lower case in order to avoid the built-in variable labels (Sx instead of SX), « x y x2 y2 xy n « x " X" TAG ' x' STO y " Y" TAG ' y' STO x2 " X2" TAG ' x2' STO y2 " Y2" TAG ' y2' STO xy " XY" TAG ' xy' STO n "N " TAG 'n ' STO x n / "" TAG '' STO y n / "ÿ" TAG 'ÿ' STO x2 n SQ * - "Sxx" TAG 'Sxx' STO y2 n ÿ SQ * - "Syy" TAG 'Syy' STO xy n ÿ * * - "Sxy" TAG 'Sxy' STO Sxx n 1 - / DUP "sx" TAG 'sx' STO ROOT "Var x" TAG 'varx' STO Syy n 1 - / DUP "sy" TAG 'sy' STO ROOT "Var y" TAG 'vary' STO Sxy n 1 - / ROOT "Cov sxy" TAG 'sxy' STO Sxx Syy / "Slope b1" TAG 'b1' STO ÿ b1 * - "Intercept b0" TAG 'b0' STO Sxy b1 Sxx * - "SSR Q" TAG 'Q' STO Q n 2 - / "S2" TAG 'S2' STO S2 ROOT "S" TAG 'S' STO S Syy ROOT / "Db1" TAG 'Db1' STO b1 Db1 / "t Obs" TAG 't' STO Sxy SQ Sxx Syy * / DUP "Det Coef R2" TAG 'R2' STO ROOT "Corr Coef R" TAG 'R' STO » »
From: leon401 on 14 Jun 2008 13:01 Hello this is the program i created few weeks ago hope it can help you « 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 Y X N DECI B0 B1 SE SE2 COVXY Vx SY SX X2 SX2 XY SXY R r YR XB YB SYR YMENOSYR YMENOSYR2 SYMENOSYR SYMENOSYR2 YMENOSE YMASE ALPHA SALPHA « Y LIST 'SY' STO @sumatoria Y X LIST 'SX' STO @ sumatoria X X SQ 'X2' STO @ Lista x cuadrado X2 LIST 'SX2' STO @ sumatoria x cuadrado X Y * LIST 'XY' STO @ Lista xy XY LIST 'SXY' STO @ Sumatoria xy SY SX2 * SXY SX * - SX2 N * SX SQ - / DECI RND 'B0' STO @ bo N SXY * SX SY * - N SX2 * SX SQ - / 4 RND 'B1' STO @ b1 B0 X B1 * + DECI RND 'YR' STO @ Lista Y roof YR LIST 'SYR' STO @ sumatoria Y roof SY N / 'YB' STO @ Y BARRA SX N / 'XB' STO @ X BARRA Y YR - 'YMENOSYR' STO @ LISTA Y - Y ROOF YMENOSYR LIST 'SYMENOSYR' STO @ SUMATORIA Y - Y ROOF YMENOSYR SQ 'YMENOSYR2' STO @ LISTA Y - Y ROOF CUADRADO YMENOSYR2 LIST 'SYMENOSYR2' STO @ SUMATORIA Y - Y ROOF CUADRADO SYMENOSYR2 N 2 - / 'SE2' STO @ SE2 VARIANZA RESIDUAL SE2 0.5 ^ DECI RND 'SE' STO @ ERROR = DESVIASION STANDARD Y SE - 'YMENOSE' STO @ Y MENOS ERROR Y SE + 'YMASE' STO @ Y MAS ERROR SX2 SX 2 ^ N / - N / 'Vx' STO @ VARIANZA SXY SX SY * N / - N / 'COVXY' STO @ COVARIANZA X XB - 'XMENOSXB' STO @ X - X BARRA XMENOSXB LIST 'SXMENOSXB' STO @ SUMATORIA X - X BARRA Y YB - 'YMENOSYB' STO @ Y - Y BARRA YMENOSYB LIST 'SYMENOSYB' STO @ SUMATORIA Y - Y BARRA XMENOSXB YMENOSYB * 'ALPHA' STO @ X-XB * Y-YB LISTA ALPHA LIST 'SALPHA' STO @ SUMATORIA ALPHA SALPHA XMENOSXB SQ LIST YMENOSYB SQ LIST * 0.5 ^ / 'r' STO @ r r SQ 'R' STO @ R SY "SY" TAG SX "SX" TAG X2 "X^2" TAG SX2 "SX^2" TAG XY "xy" TAG SXY "sum xy" TAG YR "y^" TAG XB "x barra" TAG YB "y barra" TAG SYR "sum y^" TAG YMENOSE "y-se" TAG YMASE "y+se" TAG B0 "b0" TAG B1 "b1" TAG SE "Error" TAG SE2 "Vari Res" TAG COVXY "Covar" TAG Vx "Variansa" TAG R "R coe deter" TAG r "r coe corre" TAG YMENOSYR "y-y^" TAG YMENOSYR2 "(y-y^)2" TAG SYMENOSYR "S(y-y^)" TAG SYMENOSYR2 "S(y-y^)2" TAG » » Saludos Leon
From: John H Meyers on 14 Jun 2008 15:00 On Sat, 14 Jun 2008 11:43:43 -0500, lacro wrote: > I want to make a very-very-simple application of this subject. I've > noticed I became rusty with the usage of my current HP 50g. All this > mess is due to the fact that I should get all the stat data, BUT > WITHOUT THE SDAT matrix values, although I might only provide the 6 > initial summary statistics as the local variables defined. Older HP calculators accumulated _only_ the sums (x, y, n, x^2, y^2, x*y) while the HP48 series accumulates _only_ the actual data points (from which the sums are computed when requested). An advantage of the more modern approach is more accurate results, particularly in more challenging situations. [r->] [OFF]
From: lacro on 14 Jun 2008 17:08 Tks, Leon (although you handle spanish fairly well, I keep it in english for the group policies). Very interesting and thorough approach, although couldn't test it yet or comprehend how it works at first. Tks, John (again as always). I couldn't find that very interesting distintion reference you mentioned. The disadvantage (is it so?) might be this situation, in which I need to go on from these accumulated computed sums, or in other words, the drawback is I can't make use of all the statistical tools in such a case. Best rgds, Carlos PS: There should be an economic and proper way to order all this unstructured content (*.TXT formatted) in newsgroups such as this one, in order to avoid all this repeated consults (I'm pretty sure it was already addressed). ------------------------------------------------------------------------------------- On Jun 14, 4:00 pm, "John H Meyers" <jhmey...(a)nomail.invalid> wrote: > On Sat, 14 Jun 2008 11:43:43 -0500, lacro wrote: >> ... I should get all the stat data, BUT WITHOUT THE SDAT matrix values, >> although I might only provide the 6 initial summary statistics as the local variables defined. > Older HP calculators accumulated _only_ the sums > (x, y, n, x^2, y^2, x*y) > while the HP48 series accumulates _only_ the actual data points > (from which the sums are computed when requested). > > An advantage of the more modern approach is more accurate results, > particularly in more challenging situations. > > [r->] [OFF]
From: Leon on 14 Jun 2008 19:11
I made this for my Econometric Class i only enter Stack 4 { all the values of Y } @ List 3 { all values of X } @ List 2 Number of X @ integer 1 Decimals of X @ integer for econometric porpuses i put the round RND but if you dont need you can take it out the program should work since i used it for testing my homeworks and i got all the values right if you need more help with the code please reply the email i send you because i made few other programs to solve diferent things like obly b0 and b1 Greetings Leon |