R— repeating linear regression in a large dataset

问题

I'm an R newbie working with an annual time series dataset (named "timeseries"). The set has one column for year and another 600 columns with the yearly values for different locations ("L1," "L2", etc), e.g. similar to the following:

Year    L1     L2     L3    L4
1963   0.63   0.23   1.33  1.41
1964   1.15   0.68   0.21  0.4
1965   1.08   1.06   1.14  0.83
1966   1.69   1.85   1.3   0.76
1967   0.77   0.62   0.44  0.96

I'd like to do a linear regression for each site and can use the following for a single site:

timeL1<-lm(L1~Year, data=timeseries)
summary(timeL1)

But I think there must be a way to automatically repeat this for all the locations. Ideally, I'd like to end up with two vectors of results-- one with the coefficients for all the locations and one with the p-values for all the locations. From some searching, I thought the plyr package might work, but I can't figure it out. I'm still learning the basics of R, so any suggestions would be appreciated.

回答1:

You can do this with one line of code:

apply(df[-1], 2, function(x) summary(lm(x ~ df$Year))$coef[1,c(1,4)])
                   L1           L2          L3          L4
Estimate -160.0660000 -382.2870000 136.4690000 106.9820000
Pr(>|t|)    0.6069965    0.3886881   0.7340981   0.7030296

回答2:

A combination of apply and lapply can accomplish this.

d <- read.table(text="Year    L1     L2     L3    L4
1963   0.63   0.23   1.33  1.41
1964   1.15   0.68   0.21  0.4
1965   1.08   1.06   1.14  0.83
1966   1.69   1.85   1.3   0.76
1967   0.77   0.62   0.44  0.96", header=TRUE)

year <- d$Year
d <- d[,-1]
models<-apply(d, 2, function(x) lm(x ~ year))
summaries <- lapply(models, summary)
pvals <- lapply(lapply(summaries, coefficients), function(x) x[4])
coefs <- lapply(lapply(summaries, coefficients), function(x) x[1])

来源：https://stackoverflow.com/questions/28972652/r-repeating-linear-regression-in-a-large-dataset

标签

regression