Multiple Linear Regression in R | R Tutorial 5.3 | MarinStatsLectures
Вставка
- Опубліковано 7 вер 2024
- Multiple Linear Regression Model in R with examples: Learn how to fit the multiple regression model, produce summaries and interpret the outcomes with R! 💻 Find the free Dataset & R Script here (statslectures.... We recommend you first watch the video on simple linear regression concept ( • Simple Linear Regressi... ) and in R ( • Simple Linear Regressi... ) 👍🏼Best Statistics & R Programming Tutorials: ( goo.gl/4vDQzT )
►► Like to support us? You can Donate (bit.ly/2CWxnP2), Share our Videos, Leave us a Comment, Give us a Like or Write us a Review! Either way, We Thank You!
In this R video lecture you will learn to use "lm", "summary", "cor", "confint" functions among others. You will also learn to use "plot" function for producing residual and QQ plots in R.
We recommend that you first watch our videos on the concept of simple linear regression ( • Simple Linear Regressi... ) and simple linear regression with R ( • Simple Linear Regressi... )
◼︎ Table of Content:
0:00:07 Multiple Linear Regression Model
0:00:32 How to fit a linear model in R? using the "lm" function
0:00:36 How to access the help menu in R for multiple linear regression
0:01:06 How to fit a linear regression model in R with two explanatory or X variables
0:01:19 How to produce and interpret the summary of linear regression model fit in R
0:03:16 How to calculate Pearson's correlation between the two variables with R
0:03:26 How to interpret the collinearity between two variables in R
0:03:49 How to create a confidence interval for the model coefficients in R? using the "confint" function
0:03:57 How to interpret the confidence interval for our model's coefficients in R
0:04:13 How to fit a linear model using all of the X variables in R
0:04:27 how to check the linear regression model assumptions in R? by examining plots of the residuals or errors using the "plot(model)" function
►► Watch More:
►Linear Regression Concept and with R bit.ly/2z8fXg1
►R Tutorials for Data Science bit.ly/1A1Pixc
►Getting Started with R (Series 1): bit.ly/2PkTneg
►Graphs and Descriptive Statistics in R (Series 2): bit.ly/2PkTneg
►Probability distributions in R (Series 3): bit.ly/2AT3wpI
►Bivariate analysis in R (Series 4): bit.ly/2SXvcRi
►Linear Regression in R (Series 5): bit.ly/1iytAtm
►ANOVA Concept and with R bit.ly/2zBwjgL
►Linear Regression Concept and with R bit.ly/2z8fXg1
► Intro to Statistics Course: bit.ly/2SQOxDH
►Statistics & R Tutorials: Step by Step bit.ly/2Qt075y
This video is a tutorial for programming in R Statistical Software for beginners, using RStudio.
Follow MarinStatsLectures
Subscribe: goo.gl/4vDQzT
website: statslectures.com
Facebook:goo.gl/qYQavS
Twitter:goo.gl/393AQG
Instagram: goo.gl/fdPiDn
Our Team:
Content Creator: Mike Marin (B.Sc., MSc.) Senior Instructor at UBC.
Producer and Creative Manager: Ladan Hamadani (B.Sc., BA., MPH)
These videos are created by #marinstatslectures to support some courses at The University of British Columbia (UBC) (#IntroductoryStatistics and #RVideoTutorials for Health Science Research), although we make all videos available to the everyone everywhere for free.
Thanks for watching! Have fun and remember that statistics is almost as beautiful as a unicorn!
In this tutorial, you will learn to fit a multiple linear regression model in R, produce and interpret the summary of linear regression model fit , produce residual and QQ plots in R, interpret the summary of linear regression model fit, calculate Pearson's correlation between the two variables, interpret the collinearity between two variables, create a confidence interval for the model coefficients in R, check the linear regression model assumptions and more. Download Practice Dataset: (bit.ly/2rOfgEJ); Like to support us? You can Donate statslectures.com/support-us or Share our Videos and help us reach more people!
Your videos are very insightful and easy to follow. I'm paying $$$ for R course in college, rather I could have learned by following your videos and save my money!
This entire series was just LOVELY! Thank you for providing clarity on interpreting various attributes of lm model
I agree Abhee!
Free Video on how to fit and interpret output from a multiple linear regression model in R Multiple Linear Regression in R (R Tutorial 4.12) #stats #rstats
Good series, thanks.
Glad you enjoyed them!
MarinStatsLectures Where can i find the datasets you are using
saikumar allaka You can find the dataset in the "About" section of the video, as well as the "About" section of our channel
MarinStatsLectures Thank you!! .. Can you explain more on logistic regression..
these videos have literally saved my life for every homework assignment!! Cheers from a linguistics grad student
great to hear :)
I have struggled so much in learning r. Your tutorials made learning easy. Thank you!
I have gone through all tutorials. Very clear, well organized. Very helpful.
The good part for this video is that it gives very specific details and steps in how to make a model and build up pictures in R. On the other hand, this video seems too complex about some simple parts. It discusses not very much about the background information. But as a learning guide, it seems very useful and efficient for those who just starts get used to the model in R programming.
Thanks for these videos! Saved me when I was doing my essay that required data analysis🙌
I've done most tutorials so far. They are clear and well made. Compliments!
Thanks DH1979amsterdam , and thanks for watching!
I successfully tuned linear model with a rmse 8 while watching this video. Thank you!
Mike, that help function you showed at start is so helpful as a beginner. Thank you.
You are just awesome!! I wish I had you as my econometrics professor at school! I am falling in love with econometrics and R language
it's very clear, understandable, beautiful explanation. From Japan.
Thank you! Much clearer than my professor.
Just WOW! What an amazing collection you have made! Best videos on Stats.... :)
Thanks! We'll be working on general intro stats videos over the summer, so keep an eye out for those :-)
Very useful tutorial and a wonderful, clear and concise video. What if our estimate value in a multi regression model is negative, while p-value still being significant? How do you interpret a negative estimate value for one of the independent variables? Could you please explain a bit?
Thank you so much for the great videos. I do not understand what it means when the p value is not significant in the example where age and height are in the model. But later there is a high value of correlation between age and height. Thank you.
Thank you very much!!! A quick question. Since we’re writing the hypotheses for the F-test as: H0: β1 = β2 = ... = βk = 0, is it possible to use H0:βk=0 for the t-test (instead of H0: βj=0)? Thank you :)
Hello Martin, Thanks a lot for your vlogs. They helped me a lot to get a grip on "R". Since I am new to "R", I have a basic question. I have to forecast the number of smart navigation system (Explained Variable) for several countries up to 2020. My explanatory variables are the number of cars and GDP per head in each given country. My explained variables are Austria_SNS, Belgium_SNS, Canada_SNS etc. with 3 data points for each country (2013,2014 and 2015). My explanatory variables are Austria_Cars, Belgium_Cars, Canada_Cars, Austria_GDP, Belgium_GDP, Canada_GDP etc with data from 2000 up to 2020. I would like to run the same model for all individual countries in one go replacing the name of countries with a string (something like X_SNS= X_Cars+X_GDP where X= Austria, Belgium, Canada etc). Thanks for your time in advance, Qmars
Hello there, great video! I have a question regarding when to omit a variable. On what grounds do we omit an explanatory variable for a multiple linear regression model? Also, if there is an explanatory variable that is a categorical factor will there be 2 equations for the model? Thanks for your help!
Thanks!!! It really helps my statistic final project!!
great! glad to hear you found the videos helpful.
Thanks for the useful video, Mr Marin.
@Tom Anderson you are very welcome!
Thank you. I will be using R for work, and it was a good stat refresh as well.
I look forward to more videos on more advanced topics.
you're welcome, glad you find them useful! we have more in the works, and will be producing them as we find the time to do so :-)
MarinStatsLectures
Any chance you can do a video on Random Forest?
Hi Michael Tilrico , i will add it to our to-do list, and hopefully we can get to it some time sooner than later. we have a bit of a list of other video topics on the current slate, and will work on those first over the coming summer.
how would you interpret the multiple R squared value if say height was not significant but age was.
Hi, thank you so much! So nice of you to share your understanding of stat with the public. I was running some model to explain metabolic rate as a function of 3 variables and their interactions. When i look at the summary table those variables appear not significant. Instead when i look at the ANOVA table those variables become significant. Is this normal or am I missing something? I would like to hear your perspective on this. Thanks!
Very helpful and well explained !
Thanks Cristiano Lemes !
Awesome video. Very informative. Thank you.
@Liberty you are very welcome! glad you found it helpful.
How do we report the model's overall significance using the F-test? If we report using APA:
F(2, x ) = 1938, p
Hi Marin, I wanna ask at 3:13 you said: "This is the hypothesis that the slope for Height is 0." Is that a null hypothesis? So when the p-value is below 0.05, this hypothesis is rejected right?
I am in dire need of your help. Research deadline comes around the corner.
That’s correct, Ho: the slope is 0, Ha: slope is not 0. Small o slur allows you to reject Ho and have evidence that Ha is likely true
I am trying to do class() but it says object not found even though it shows a summary of the dataset. What can I do? Please helpp
Thank you for your great videos! I am looking for your video where you mean centre height and age. I am not sure if I missed it but I have watched all the videos up to 5.12 and was not able to hear where you mention it.
could you do a video on logistic regression? and choosing variables to make-up the model? this video was extremely informative!
Hi Lexi Rene , thanks! We're planning on extending the series to include Logistic regression, Poisson regression, etc, but that will take some time. You can fit a logistic regression in R using the command: *model
Thank you so much. its very helpful . please explain the multivariate regression analysis
Hi, this question is way too large to be able to answer here in a comment section. if you had a specific question id be happy to try and answer that
Hi Martin, In this video you have more independent variable, What if we have more dependent variables (e.g. three dependent variables), what kind of statistical analysis can be performed on such type of data ?
Thanks,
Faray
You are really good , thanks for your videos.
Thanks a lot,
You're great teacher.
you're welcome +Rin Sa , we appreciate that!
These are all great and very informative videos! Thank you very much for posting them. Still i am stragling a bit, I would like to check the correlation of let say 8000 genes in 10 different patients that I organized from less sick to more sick (lets say percentage of neuronal loss). There is any way I can get the correlation of the expresion of these genes with the sickness of the patient looking at the r? (ex. 0.99 would be a gene that correlate better with a worse patology, right?) if so, Can I check all the r for each gene using R?
Hi Marin, Thankyou for the wonderful videos. You have got great lecture delivering skills and your voice sounds nice. I would request you to prepare few videos for logistic regression(Binary and Multinomial) having both numeric and categorical predictors. Interpretation of coefficients, odds ratio, Graphics. Thanks
thanks for your comment +Imran Khan , we plan on adding videos for logistic regression (and other generalized linear models) when we have time to create them. it is a topic that is high up on our priority list, although finding the time to write, record, and edit the videos is the challenge!
Thank you, very helpful.
I have a question regards F-statistic. As you know when a linear model has no intercept, it assums as a nls in R. So we won't have F-statistic in the summary of model. How can we calculate this statistic in R for a no-intercept linear regression?
Hi. How can I manipulate the input's variables data on MLR to get the higher adjusted R-Square
hi, you are talking in min 2.13 about centering the x axis (here age and hight) for a better/ different intercept - where did you explain this? would be wonderful to get help with this! thank you
The link in the description to the script and dataset are broken - I was able to find the LungCapData2 file on the site, but do you have a link to the script in this video?
Great Video! What software and hardware are you using to do it? I imagine the video is separate the nice highlighting? Also what resolution it looks HD but plays back very nicely on my moderate internet connection? Thanks.
How do you identify the SSE, SSR, and SST values from these outputs?
hi . after creating the model i want to give input for each variable, since height and age are numeric they are given as numeric itself, BUT how to give input for gender,smoke,Caeserean since these are strings?
Hello ! These video clips really helped a lot ! I want to ask a question about how to use R to deal with linear regression problems contain absolute values as following ?
min 2x1 + 3|x2 − 10|
s.t. |x1 +2|+|x2|≤15.
my Goooood you guys are awesome. thank u thank u thaaaaank you.
You’re welcome :)
Hi
I've fitted a regression model to a data that has one indicator variable: 1 and -1. My question is how can I exhibit the equation of y when the indicator variable is 1 and when -1.
Amazing video, sir. Can you please help me out on how to find the Explained sum of squares of the model without using the ANOVA table? I know how to calculate the Residual Sum of Squares, but not the explained sum of squares. Thank you very much in advance.
Hi Syafiq Din , sure. im not sure why you dont want to just read it off the ANOVA table? (ie) *anova(model)*, and look at it directly from there. but, you can work it out from the model summary if you like, using the fact that *R-square = 1 - (ResidualSS/TotalSS)*, and *TotalSS = ResidualSS + ModelSS*
@3:00,.... Stats question, regarding the t-statistics for Age and Height (ignoring that the intercept is technically at -11.74). Does this mean technically that because the t-value for Age and Height are 3.45e-12 and
Hi +lilyhwangathotmail , sort of....the general idea you are getting at is correct, although there are a lot of little 'errors' in the way you're phrasing things. it's the p-values that are 3.45e-12 and
Hi. Is there any way to perform multiple linear regression on raster time series images?
Very helpful tutorials! Thanks for helping me understand R better.
Do you know how to cluster standard errors by a certain variable easily? I struggle understanding most of the answers I can find..
Hi, I'm n to exactly clear on what you mean by "cluster SEs by a certain variable"...if you can clarify that for me, i may be able to offer some suggestions
if t value of age is negative and also pvalue is also small when what should we with variable age in our linear model?
If the p-value is small (the t-stat is large, in absolute value), then this tells you that the variable age is a significant predictor of the outcome (Y), and would probably be kept in your model. of course, it depends on the goal of your model (if it is a predictive model your building, or a causal model, etc), but in general, the small p-value tells you that age is a significant predictor of lung capacity.
Mike, is there any command where I can use that gives me a summary of my independent variable in relation with each possible combination of independent ones, so I can examine over fitting and multi col ? Also does it give mallows cp and vifs? I don't even know how to express this so to google it, but I saw a video of someone in minitab doing that so I am pretty sure R should have that option
Could you explain a little bit more about the collinearity? Like in this example, since height and age are highly related, why not just use one to explain the Y?thx
hi +cherry d , it is too difficult to explain a complicated concepts like that in a discussion thread. but a short answer is that IF age and height are VERY highly correlated, we would/should use only one in the regression equation (if they contain almost the same information). if they are just fairly highly correlated, but not collinear (not nearly inseparable) then we can use both of them. id suggest to read more about collinearity (and things like VIF=variance inflation factor, Tolerance,...)
Love this..cleared my doubt as well!!
Getting r squared 0.0000845 is good or bad. Please help
how do i find the y-intercept, when I need a certain value of age for example?
Hello, I have a question. What is the difference between using regression diagnosis model and shapiro.test to figure out the normality of distribution?
Thanks.
Hi +Yating Li , they're just two different ways of trying to do the same thing. the diagnostic plot is visual, and more open to interpretation. the Shapiro test is a formal hypothesis test (with Ho: its normally distributed). worth noting is that failure to reject the null is not the same as proving the null...
very helpful tutorials
who do you write this up in a results section?
Fantastic videos! If it weren't for this channel I would not be doing so well in my Biometry class. What about when you have continuous variables on both the X and Y? Do you have any videos where you check the assumptions of MLR with continuous data? I have a homework data set that I'm working through and when I run the autoplot command, I see on my Residuals vs Fitted plot that there is a slight wedge shape that increases along the x axis. My Q-Q plot looks good with only three outliers and my Cook's D plot looks good and shows the three outliers with values under 0.6. I'm just not sure if I need to transform my data or not.
good to hear you're finding our videos helpful +playthedoghouse ! sure, we have a video on checking the assumptions of linear regression. you can find it here: ua-cam.com/video/eTZ4VUZHzxw/v-deo.html
Great video. Just wondering, how do you find the residual sum of squares and estimate of the error variance (if possible) based on this summary output?Thank you!
Hi Mark Goh , sure you can find those in the R output. Below the model coefficients you can see the "Residual Standard Error". The error-variance (which goes by many different names, like MSE, etc) would be the square of the Residual Standard Error.
You can also use the output to work out what the Residual SumOfSquares is. You can take note that the Residual Standard Error = sqrt(ResidualSumOfSquares/DFresidualSumOfSquares). You can find the DF for the residual sum of squares right next to the Residual Standard Error in the R output.
You can also just ask for the ANOVA table for the model using *anova(model)*, substituting in the name of the object you saved the model in. This will return the ANOVA table with all of the sum of squares, mean squares, etc.
What if you use Gender instead of height? How do you fit separate linear regression model for each gender?
if you include Gender, you will get a model coefficient for gender, which essentially creates 2 separate lines, one for males and one for females.
THUNK U FOR UR VALUABLE INF.
you're welcome
Very helpful! Thank you!
you're welcome +Anony Mouse
How do I perform a MLR with a dataset that contains continuous AND categorical variables? Or could anyone point me to a video explaining this
This video shows exactly that. Age and height are numeric, smoke, gender, and caesarean are all categorical variables
Hi Mike,
i wonder what smoker means in this case?
Bcs we got the age given from 0 years to 25 years. And i dont think children below 16 years should smoke/ arent allowed to.
So does Smoker=yes means that the women was smoking during her pregnancy?
Or where is that data from?
Hi +Dopserados , the Smoke variable is an indicator if they identify them selves as smoking 0=no, 1=yes. the ages are 3-19. the data is simulated data from a real dataset from the 1970s, collected in. the youngest smoker in the data is 10. while children under a certain age are not allowed to legally buy cigarettes, some still do get cigarettes and do smoke. there are, of course, all sorts of issues with self-reported smoking status, and a simple yes/no answer to this question. these are some of the things i discuss with my class about this dataset. hope that clarifies some things for you.
Thanks +MarinStatsLectures for the explanation, i will use your data for a little seminar paper and needed some information about the data. Everything is fine now. :)
Great video!
thanks +Nev Dog !
Thank you for the video. So, the "summary" argument in R gives an evaluation for a linear regression, but what about a non-linear regression model? How can we evaluate our non-linear regressions?
Hi Manizheh Rajabpour , i'd say that it gives more of a summary than an evaluation, as the term evaluation can mean many things, and this gives more of just the model coefficients, and a few other basic summaries. there are other ways to evaluate the fit of the model, etc.
for a non-linear model, it depends exactly on which command you are using, but in general, the "summary" command will return a general summary for most objects in R, and will determine what type of a summary is appropriate for the object. for example, typing summary(LungCapData) will return a summary of the dataset.
Hi MarinStatsLectures, thank you for the response. I meen summary(model) gives us some statistics of the linear model like "Residual standard error", "multiple R-squared", "F-statistic". p-value could explain the significance of our model. right? Now what about a non linear model?
In fact I have some multiple regression models that are linear but since they do not have an intercept, we have to write them as below in R:
model1 = nls(Y ~ a*X1+b*X2, data = trainDat, start = list(a = 1, b=1))
As you see it is a linear model but we write it as a non linear model, and the summary(model1) do not gives us the statistics.
What is solution for this? Or how can I evaluate the fit of both multiple linear and nonlinear models in R, if there is other ways than a summary of model?
thank you in advance
Do you have a tutorial for quantile regression and nonparametric quantile regression? If not, can you make plis?
You said collinearity would be discussed in later videos, could you tell me which one exactly?
Hi STEPHCRLR , we're still working on the linear regression series, and haven't made that video yet.
What if I have a column with a unique ID for each person (let's say the column is called UID). How would I get R to run a linear regression for each UID and save the coefficients for each person. So if I have 200 people, I would have 200 rows of data represent the coefficients for each person?
Hi +Dave , speaking in general, if you wanted to apply some function to the rows of a dataframe, you can use the *apply* function, e.g. *apply(x, MARGIN=1, FUNCTION)*....this would apply a specified function to the rows (margin = 1) of the dataframe x.
it is worth mentioning that in order to fir a regression to each individual, you must have multiople datapoint observed for each of the variables.
cool explanation man :D
Please let me know if I have in data 2 dependable variables and 16 in dependable variable , how to generate model? shall I delete one dependable variable and build the model? then vise a versa ?
it really depends on what you are trying to model, and why, etc. one option is to model "y1" and then separately model "y2". whether or not this is a good solution depends largely on the context of whay you are trying to accomplish, and why. but sure, it is not unreasonable (as a general statement) to model there two different outcomes/dependent-variables separately
Hi Martin, I would like to know how to read multiple CSV files into multiple data frames by using a loop function. I'm having various CSV files i.e., user1.csv to user100.csv and I want to read all of them into different data frames so that I can access them individually. In every user data I've to analyze some columns like Skin temperatures, Heat flux like that. It would be appreciated if you explained in the code. Thanks!!
Hi suraj kumar . you can use the *paste* command to paste a 1,2,...100 at the end of the word user, when reading in the data. here's an example, you can modify to suit your purpose:
for (i in 1:10){
print( paste("user",i,".csv", sep="") )
}
you can see that this creates a "user1.csv", "user2.csv" and so on...
you can remove the print from my code...that's just there to have it print on the screen so that you can see that it's doing what you want. you can incorporate this into the loop that has a read.table command in it to read in the data...
ps. my name is Mike Marin, not Martin...
Sorry for asking a dumb question, but is the Residual Standard error output the SSR?
it's not a dumb question,... and no, they're related, but not the same thing.
the Residual SE from output is = square-root( SSR / n-k ) , where n-k is the number of observations minus the number of parameters in the model including the intercept....in case of simple linear regression it would be n-2.
MarinStatsLectures- R Programming & Statistics thank you! Your videos are awesome!
hello Marin
could u share something about Comparison of two Population Proportions, i need to find a test to determine the difference between two proportions, in 2 differents samples. thanks...
Hi Lizbeth Hernandez , you can use the *"prop.test"* command in R, to compute a difference in two proportions test. This test tests the Null hypothesis Ho: p1 = p1. You can get this done in R using *prop.test(table(X,Y))*, where X and Y are both categorical variables with 2 levels, and you would like to calculate the proportions of Y, for each of the categories of X.
To learn about changing default values for the test, you can look at my video on the 2-sample t-test here: Two-Sample t Test in R: Independent Groups (R Tutorial 4.2) this will show you how you can change to a 1-sided test, change the confidence level, etc
Sir I wanna know before we perform multivariate regression using lm function do we have to scale the data for various variables in order to bring them in same range or lm has inbuilt scaling function ?
Hi +Aditya Kumar Singh , you can use the *scale* function in R to scale or standardize a variable, and use this in a regression model IF this is what you want to do. it is worth mentioning that in a general case, there is no need or reason to need to scale a variable in order to fit a regression model. scaling a variable may allow for more meaningful interpretations, in certain circumstances, but for the most part there isn't really a reason or need to do this.
Hello, I have a question. What do the "Residuals" for min, 1Q, median, 3Q and max mean? Are those values pertaining to the predicted values of Lung Capacity?
Hi +Sony M , the residuals are the difference between the observed y-value and the predicted y-value from the model (the vertical distance between each of the observations and the regression line). R is returning a summary of them...what is the smallest residual, the Q1 for them, and so on...
+MarinStatsLectures Thank you, really helped!
MarinStatsLectures thank you for the tutorial but could you tell me why the summary didn't show all of the level variable? like smokeyes and smokeno?
Hi Bram Setyadji , that's because the "no" is the reference, and is captured by the intercept term. the intercept is the estimated mean Y value for all X=0. in the case of the categorical variable for smoking, it is coded as no=0, yes=1, and so the non-smokers are part of the reference/intercept. in the example at the end of the video, the intercept refers to age=0, height=0, smoke=0=no, gender=0=female, and caesarean=0=no. i have a separate video that explains more about categorical variables in a linear regression, and how dummy/indicator variables are used for these. you can check that out if you need a more in depth explanation of how categorical variables work in a linear regression model: ua-cam.com/video/2s8AwoKZ-UE/v-deo.html
Thank you for the prompt reply. I've watched almost all of your videos (especially what you've mention above) and now I understand why in multi linear regression or GLM summary the reference variable is not shown.
great to hear Bram Setyadji
Hi, I have a question that may not be related to this video, but I would be appreciated if you answer me if you know. Is it logical to do ANOVA between the result of a model (regardless the kind of model including linear, nonlinear, parametric, nonparametric, ...) and true measurements to see the significance of the model?
Hi Manizheh Pourrahmati , i'm a bit unclear on what you mean by comparing results of a model and the true measurements. linear models do return an ANOVA (and F-test) for overall significance of a model, which is sort of what lie you say...but I'm a bit unclear on the wording you are using. in general, you would use ANOVA for a linear regression model, but not for all models...for generalized linear models, instead of ANOVA, you would use a likelihood ratio in place of the ANOVA.
regardless of the confusion in your question, the short answer is that no, you wouldn't use ANOVA for any sort of model...it wouldn't be appropriate for a non-parametric model, generally not appropriate for a non-linear model (depending of course on what type of non-linear model, as that term captures a ton of different sorts of models)
MarinStatsLectures, Thank you for your response. To be clear, I am building models based on a dataset and validating them using 5-fold cross validation in R. the meaning of result of model is predicted values based on test data which as a consequence of using k-fold cross validation all data are used as both training and test data and finally we would have predicted data. my models are multiple linear and non-linear regressions and also neural network and random forest. As you said f-test for linear regression gives us the significance of the model. what about other kinds of model? I thought maybe by comparing predicted values against real values (in situ measurements) and based on anova, we can explain the significance of our model. If it is not logical, which method is used for expressing the significance of models except for linear regressions (random forest model, neural network and non linear Reg.)?
Hi Manizheh Pourrahmati , thanks for clarifying. i wouldn't consider myself an expert in neural networks or random forests, so i don't think i could offer any better advice about those then you will be able to find in other places. but i can say that in general, it sounds like you will want something other than the significance of your model (a significance test of you overall model would just tell you if your model is better than nothing). it sounds like you want to compare competing models, to decide which is better? for that, you may want to look into things like AIC/BIC to compare models that are non-nested (one does not have to be a subset of the other). the only thing for these though is that it is important that your Y variable is on the same scale for the competing models, in order to use something like AIC to compare models (e.g.) if one model uses Y and the other uses ln(Y), then not good to compare models using AIC or BIC
HiMarinStatsLectures Thanks a lot for answering my questions and trying to solve them even they are not pertinent. In fact I developed different kinds of models and used R -squared adjusted and RMSE to find the accuracy of my models and also AIC to compare the models. I wrote a paper based on my result. I received following comment from a reviewer:
"Regression analysis generates an equation to describe the statistical relationship between one or more predictor variables and the response variable. To determine whether a result is statistically significantly different, authors have to perform further analysis and then compare the model (optimum model). Statistically, present analysis does not take into consideration such teats."
I do not know what exactly he wants, but I thought maybe he means the significance of models. Do not you think so?
Hi Manizheh Rajabpour , it a bit tough to tell because i don't know exactly what you had written in your paper, that the reviewer's comment is referencing. but it sounds like they might be suggesting that if you are using the adjusted R^2 to compare different models, that you might want to do a significance test, to test if the adjusted R^2 are significantly different, and not just that one is greater than the other. but like i said, it's tough to tell without knowing exactly what you had written in your paper that they are referring to.
amazing!!
what is the meaning of e in the p-value
it means 2.2 *x 10^* -16
How could I do multiple linear regression if I have thousands of variables
Also, could you teach me how to do the quartile regression as well
thx
hi i know this is old, but its about your categories of data not the number of data points, ie height vs weight, or for multiple regression, height vs weight vs strength.
where can i get this dataset
what about multiple linear regression simulation steps?
Hi +Fatma Ahmad , I'm not sure exactly what you are trying to ask. can you clarify your question, and i can try to help?
+MarinStatsLectures ,
hi ,thanks for your reply,
I need to generate data follow the terms of the least squares method and multiple linear regression model-simulation- then i will study multiple linear regression analysis and multiple fuzzy linear regression analysis.
Hi Fatma Ahmad , i'm still not completely clear on what you mean. i will provide an answer here, and hopefully this is in line with what you are looking for. below, is some code (that you can modify) that will randomly generate an "x1" an "x2", and then generate a "y" using these, and adding some error to it.
# randomly generate 100 observations from a normal distribution with mean 0, sd=1
*x1
+MarinStatsLectures
thanks alot for your help , that's exactly what i want :) .
+MarinStatsLectures
excuse me , I have another question? can I transfer the outputs of x1, x2 and y into another statistical program sush spss , minitab or matlab?
can't we use "pairs()" command
Yes, you can use the "pairs" command to produce all possible pair-wise scatterplots of the variables in your data set.
how many bananas is the lung capacity measured in ?
Really now is it US or metric system?
i can't recall the units of measurement...but it is a measure of volume displacement per unit time