stepwise regression: Undefined function ' stepwiselm' for input arguments of type 'cell'.

I have one .txt file and converted to first a table Ta(Ta=readtable（'xxx.txt'）) then an array Aa（Aa=table2array(Ta)）, the .txt file contains 220 rows and 12 cols, but the table and the array only have 219 rows and 1 col. Where did I do wrong?

Then when I tried to do stepwise regression I got error message: Undefined function ' stepwiselm' for input arguments of type 'cell'.

My coad was: mdl=stepwiselm(Aa)

In the .txt file, the first raw are texts e.g. elevation, hight, yields etc. I though I could use these names to define Predictor variables and Response variable. But since these names are lost in Aa, how should I write code for stepwise regression?

[Roger Wohlwend]

Please attach the file to your question.

[Yaojiayin]

exercise_data333.txt(220*10) contains data for sample points. I converted it to a table and then an array A. Then use A in the code for calculation to get a new .txt called A-100spreg2-raa12.

I then converted the new .txt to a matrix M M = dlmread('A-100spreg2-raa12.txt')

Then I want to apply stepwise regression on M.

In exercise_data333.txt, the first row are texts, but in M there are no text. Then how can I write e.g. use 'yield' as dependent variable i.e. response variable and 'DEM, grad, HI' as independent variables i.e.predictor variables?

• 'yield' is the 10th colmun, so stepwise(M,M(:,10)) works fine, but what if I want to keep names for all variables.

Answer 1

[Roger Wohlwend]

I recommend using a dataset for your problem.

A = dataset('File','excercise-data333.txt');

A dataset is a kind of table. The headers do not get lost and you can easily address a column (for expample: b = A.yield or c = A.x). The function stepwiselm accepts datasets as first argument.

[Yaojiayin]

Instead of exercise-data333, I used another .txt file called Araa06.txt, which contains the same variable names as exercise-data333(first row) but new values (after ran the code). Then I ran A = dataset('File','Araa06.txt')

Two errors returned:

Misuse dataset/readFile (line 207) Variable lengths must all be the same.

Error dataset (line 351) a = readFile(a,fileArg,otherArgs) What do they mean?

[Roger Wohlwend]

Hard to tell. The function can handle missing values, so that cannot be the problem. I have no idea.

[Yaojiayin]

For some reason it works now! (I simply copied the first row(names of variables) form the original file exercise-data333.txt to Araa06.txt, then created dataset A, then ran stepwise regression.)

Thank you so much!