Changing X and Y variable for linear regression

Question

I am currently looking at if there are annual trends in my data. I am doing linear regressions between R and year, and H and Year, as well as between R and H.

However, when I do a linear regression of R against year where year, I get a NA F and P value. When year is y the code works, but I would like to know why the linear model only works one way, and if year as y is valid in this instance for data analysis? Thank you in advance.

DATA:

year	R	H
2000	160	140
2001	178	153
2002	149	138
2003	161	149
2004	180	173
2005	150	142
2006	158	130
2007	149	190
2008	167	200
2009	172	204

Code:

#this has lots of NA outputs
linearmodel<-lm(data$R ~ data$year)
linearmodel
summary(linearmodel)

#this gives output statistics
linearmodel<-lm(data$year ~ data$R)
linearmodel
summary(linearmodel)

Thank you again.

If I construct the dataframe from scratch with the same data it doesn't give me problems, maybe the issue is in the dataframe you are working on. Are you sure the values referring to the years are numeric and not strings? — Giulio Mattolin
What is the output of str(data) ? If year is a factor then it is not possible to run lm with year as dependent variable — Basti
Thank you for trying to replicate. The output of the str is that each row is numerical but year has this... $ year: chr [1:10], I converted the dataframe into a new csv file and now have no errors. I assume the issue was with the dataframe as it had been rearranged and formatted a number of times. Thanks again both! — Emma

Gregor de Cillia Gregor de Cillia · Accepted Answer · 2021-07-27T12:22:14

I can't reproduce this. Both formulas contain no NA values in the output. The same is true for the summary() of those two models.

data_68544559 <- data.frame(
  year = 2000:2009,
  R = c(160, 178, 149, 161, 180, 150, 158, 149, 167, 172)
)

lm(R ~ year, data_68544559)
#> 
#> Call:
#> lm(formula = R ~ year, data = data_68544559)
#> 
#> Coefficients:
#> (Intercept)         year  
#>   259.58788     -0.04848
lm(year ~ R, data_68544559)
#> 
#> Call:
#> lm(formula = year ~ R, data = data_68544559)
#> 
#> Coefficients:
#> (Intercept)            R  
#>   2.005e+03   -3.316e-03

Changing the format for the time column to POSIXct will also give no NAs, but an error in summary() because the residuals can't be calculated.

summary(lm(ISOdate(year, 1, 1) ~ R, data_68544559))
#> Error in Ops.difftime((f - mean(f)), 2) : 
#>   '^' not defined for "difftime" objects

Changing X and Y variable for linear regression

1 Answers