NEWS

---

jstats 0.9.267

Thank you for trying jstats. Here is what is new since 0.9.265.

• "Not at all" scales read as Likert. A single question whose labels run from Not at all to Extremely, Very or A great deal, among others, now reads as a Likert item: jscreen() shows it as Likert, and jdesc() no longer warns that it seems categorical.
• Infinite values. When a variable holds Inf or -Inf, such as the log of a zero, jt(), jaov(), jlm(), jlogistic(), jalpha(), jcorr() for Pearson's r, and jplot() for a histogram, scatter or box plot say how many, and give the line that sets them to missing in a new variable. Before, R stopped with "missing value where TRUE/FALSE needed" or "NA/NaN/Inf in 'y'". jdesc() prints -Inf with its sign, where it printed Inf.
• Messages easier to read. Notes, warnings and errors put one sentence on a line, and they say outcome and predictor where they said DV and IV.
• Words beyond A to Z in jencode(). When the words hold letters such as Á or 北, its note says it coded them in character-code order, which may not be alphabetical: Ávila comes after Zebrinha. A map sets any order you like: d$CityN <- jencode(d, City, map = "Ávila=1; Belém=2; Zebrinha=3")

When something you typed before now stops, its message says what to type instead.

Problems or suggestions: tell me, or open an issue at https://github.com/jma61/jstats/issues.

jstats 0.9.265

Thank you for trying jstats. Here is what is new since 0.9.254.

• Chinese, Japanese and Korean text. Tables, notes and labels line up, each character taking two columns. Text loaded from a CSV file sorts, so jencode() codes it, where it stopped with an encoding error. A full-width semicolon or equals sign typed in a map stops, with the call retyped in plain characters, ready to run.
• Labels survive a subset. A subset taken with base R, such as community[community$Age > 40, ], keeps its value labels and declared missing values from the moment jstats is loaded. Before, a subset taken ahead of your first jstats call lost them without a word, and -99 (Refused) was counted as an income.
• A data frame's name typed wrong. Type dd for d, as in jt(Score ~ Group, dd), and jstats offers the call with the data frame that holds those variables, jt(Score ~ Group, d). The same inside an expression, such as dd[dd$Age > 40, ].
• Blank text cells in Excel sheets. A sheet's text variable now loads as a CSV file's does: an empty cell is the <blank> category, counted as valid, where it loaded as missing. To keep blanks missing: jload("survey.xlsx", na = "")
• More single questions read as Likert. Labels running from Never to Very often, Very poor to Very good, or Strongly oppose to Strongly support, among others, now read as a Likert item: jscreen() shows it as Likert, and jdesc() and jcorr() no longer warn that it seems categorical.
• Plots of labelled variables. A labelled variable with a few categories is drawn as categories, unless numeric = says otherwise. So jplot(WellbeingScore ~ Region, community) draws a boxplot for each region, where it plotted the region codes as numbers.

When something you typed before now stops, its message says what to type instead.

Problems or suggestions: tell me, or open an issue at https://github.com/jma61/jstats/issues.

jstats 0.9.254

Thank you for trying jstats. Here is what is new since the last release (0.9.252).

• CSV files load as written. A postcode such as 0800 keeps its zero, a long ID keeps every digit, a column of F stays F, and the word NA in a text column (Namibia's country code) stays a word. A note says when a column of numbers was kept as text, so a file you loaded before may now give such a column as text, with the reason beside it.
• N/A, - and . in a column of numbers. jstats does not assume they mean missing. When a few such cells make a column of numbers text, a note names them and gives the line that reads them as missing: jload("incomes.csv", na = c("N/A", "-", "."))
• Keep a column exactly as typed. text = reads codes such as 250.10 as written, in CSV and Excel files: jload("visits.csv", text = Diag)
• What is new, when you update. From your next update on, jupdate() prints these notes for each release it installs.

When something you typed before now stops, its message says what to type instead.

Problems or suggestions: tell me, or open an issue at https://github.com/jma61/jstats/issues.

jstats 0.9.252

To the colleagues and students trying jstats before its CRAN release: thank you. Here is what is new since the last release (0.9.223).

• Quiet mode for scripts. joutput("silent"), or print = FALSE on any analysis, fits a model or builds a table without printing it. Keep the result (m <- jlm(...)) and use it later.
• Label a set of survey items in one call. jrelabel(d, c(Q1, Q2, Q3), labels = "1=Never; 2=Sometimes; 3=Often", modify = TRUE)
• New plots. type = "density" beside the histogram, facet = for one panel per group, and the log-odds plot of a logistic regression, jplot(m, which = "logit").
• More in the model output. Standardized coefficients in jlogistic(), Cramer's V in jcrosstab(chisq = TRUE), and one VIF row for a categorical predictor.
• Missing values on text variables. jdeclare_missing() now declares codes such as "UNKNOWN", or blank cells, on a text variable.
• Loading files. jload() reads the package's example files by name, skips title rows in Excel and CSV files (skip =), reads files with no names row (header = FALSE), and fixes variable names R cannot read, keeping the original as the label.
• Clearer messages throughout. Every line of code a message suggests now runs as printed.

Problems or suggestions: tell me, or open an issue at https://github.com/jma61/jstats/issues.