Paleo-Balkanic *mendjo- 'foal' (updated)

Albanian mëz, mâz, Romanian mînz (a Dacian substrate loanword) 'foal' derive from a Paleo-Balkanic¹ protoform *mendjo-² related to Gaulish mandu-, borrowed in turn into Latin mannus³. Also related are Basque mando 'mule' (dialectally also 'sterile animal or woman'), presumably a Celtic loanword, as well as dialectal High German (Tirol) Manz, Menz 'sterile cow'⁴ and Italian manzo 'ox'⁵.

Basque idi 'ox' can also be derived from this etymology, assuming the following changes (not necessarily in that order): 1) loss of m-, 2) e > i, 3) nd > d, 4) -o > -i'⁶.

Probably also belong here Sanskrit mandurā 'stable for horses', mandira- 'dwelling, house' and Greek mándra 'pen, stable'. In my opinion, we're dealing with a Wanderwort of ultimate Altaic origin: Tungusic *manda-ksa 'Eurasian elk (Alces alces)', Mongolian *mandʒi 'male elk' and Turkic *buto 'young of camel'  (EDAL 1253). 
_____________________________

¹ Certainly not IE-native as thought by Mallory-Adams. See J.P. Mallory & D.Q. Adams (2006): The Oxford Introduction to PIE and the PIE World, p. 142.
² T. Gaitzsch (2010): Das Pferd bei den Indogermanen: Sprachliche, kulturelle und archäologische Aspekte, p. 263-264.
³ X. Delamarre (2008): Dictionnaire de la langue gauloise, p. 214.
⁴ J. Pokorný (1958): Indogermanisches Etymologisches Wörterbuch, p. 729.

⁵ Possibly a Messapic loanword, cfr. Menzana, a divinity to whom horses were offered.

⁶ In sharp contrast with more recent loanwords, the native core of Basque underwent so many phonetic changes, up to the point of making it hardly recognizable.

The ancient Basque homeland (updated)

According to most specialists, the area located between the Garonne river and the Pyrenees, ancient Aquitaine and modern Gascony, is the homeland of the ancient Basques. There we find a high density of Gascon toponyms in -òs, which extends to Basque -oz(e), -otz(e) and Navarro-Aragonese -ués on the other side of the Pyrenees¹.

This toponymic element derives from Celtic *ouxsV- 'high' (Old Irish úais, Cornish a-ūch)², whose superlative *uxsV-(s)amo- 'the highest' can be found in Middle Welsh uchaf, and the femenine *uxsV-(s)amā in Gaulish Uxisama (modern Oisème), Uxama³ and Celtiberian Usama (modern Osma). On the other hand, the former proposal of the Spanish linguist Ramón Menéndez Pidal, who linked the toponymic element to Basque otz 'cold' (itself from Celtic *ouxtu-⁴), can be dismissed.

The same lexeme would also be part of the Aquitanian anthroponym element Andos(s)-, Andox- (Latinized as Andossus, Andoxus) 'lord'⁵, whose first member would be the Gaulish intensive prefix and- 'very'. In my opinion, this evidence, together with loanwords such as gizon 'man' < Gaulish gdonio-, would indicate a Celtic substrate in Paleo-Basque whose existence hasn't been yet discovered by academic Vascologists.

________________________

¹ G. Rohlfs (1970): Le Gascon. Études de phylologie pyrénénne. p. 29-33.

² R. Matasović (2009): Etymological Dictionary of Proto-Celtic, p. 303.
³ X. Delamarre (2008): Dictionnaire de la langue gauloise, p. 329.
⁴ R. Matasović, op. cit., p. 304.
⁵ Replaced in modern Basque by jaun < *e-aun, presumably a fossilized participle 'elevated'. 

Man and dog (updated)

Uralic *koje 'man, person' is a remnant of a Borean word which spread as a Wanderwort into East Caucasian *χχHweje (NCED 26) and Tibeto-Burman *qhʷi:j 'dog'¹. The suffixed variant  *koj(e)-ra  > Uralic *kojra 'male (dog, man)' would correspond to East Caucasian *χχHwej-rV 'dog' (oblique stem) and Kartvelian (Svan) xwir- 'male (dog)'. IE *wi:r-o- 'man, husband'² would also belong here, although probably as an inherited word. 

On the other hand, Sinitic *khʷi:-n 'dog' borrowed into IE *k´(u)wo:n (a cultural loanword whose direction has been often reversed by Indo-Europeanists³) would be derived from the same lexem with a different suffix. 

Unfortunately, most macro-comparativists are unable to differentiate between borrowed and inherited lexicon⁴, as they blindly apply the comparative method coined in the 19th by Neogrammarians, which assumes common inheritance from a single source, represented by the genealogical tree model. By applying it to some hundreds of words, this process ultimately leads to the reconstruction of non-existent macro-families whose chronology is shallower (typically 2-3 times) than the actual ones. 
______________________ 
¹ For which Starostin reconstructed a Sino-Caucasian etymology.
² For which Starostin reconstructed an Eurasiatic (Nostratic) etymology.
³ T.V. Gamkrelidze & V.V. Ivanov (1995): Indo-European and the Indo-Europeans, p. 507. 

⁴ Generally, the lesser the semantic latitude, the likelier we're dealing with a Wanderwort, as it's often the case with names of domesticated animals.